Methods for identifying polypeptides

Abstract

The present disclosure relates to methods and systems for identifying polypeptides using nucleic acid data obtained from a polypeptide sequencing device that encodes amino acid sequence information as barcoded nucleic acid sequences that are subsequently analyzed in a nucleic acid sequencer. The nucleic acid data generated as output from the nucleic acid sequencer is processed to extract information about the analyzed polypeptides. The process includes, for the generated nucleic acid data, generating a plurality of concatemer identifier strings; and providing the plurality of concatemer identifier strings as input to a computer model configured to infer an amino acid sequence of a polypeptide from a concatemer identifier string. The present disclosure has utility at least in a variety of methods and related systems for high-throughput polypeptide analysis and sequencing.

Description

[0001] Cross-reference to related applications

[0002] This patent application claims priority to U.S. Provisional Patent Application No. 63 / 597,668, filed November 9, 2023, the disclosure and contents of which are incorporated herein by reference in their entirety for all purposes.

[0003] Reference to electronic sequence listing

[0004] The contents of the electronic serial number (776532005240SEQLIST.xml; size: 35,043 bytes; and creation date: November 6, 2024) are incorporated herein by reference in their entirety. Technical Field

[0005] This disclosure generally relates to biotechnology, and more particularly to a method for identifying peptides using nucleic acid sequence data obtained from a peptide sequencing device that encodes amino acid sequence information into barcoded nucleic acid sequences. The disclosed method is practical in a variety of high-throughput peptide analysis and sequencing applications. Background Technology

[0006] High-throughput sequencing of peptide analytes in biological samples remains a challenge. Several methods for high-throughput peptide sequencing have been published, including US 9435810 B2, WO2010 / 065531A1, US 2019 / 0145982 A1, US 2020 / 0348308 A1, US 20200209255 A1, US 11549942 B2, US 20180299460 A1, US20210079557A1, and US 2023 / 0136966 A1, which utilize sequential amino acid or epitope recognition by a limited set of degenerate binding reagents as a key step during peptide sequencing assays. Furthermore, reagents for cleaving components of peptides recognized by binding reagents have been employed or developed independently (see, for example, US 11427814 B2). However, the inherent difficulty in distinguishing structurally similar individual amino acid residues within peptides leads to high levels of errors, making the decoding of output data from peptide sequencing a challenging task and hindering the widespread adoption of high-throughput peptide sequencing methods. Further improvements in techniques related to decoding high-throughput peptide sequencing data remain needed in this field.

[0007] This disclosure describes algorithms for processing peptide sequencing data to meet the aforementioned and other needs. These and other embodiments of the invention will become apparent upon reference to the following detailed description. For this purpose, various references are set forth herein, which describe certain background information, procedures, compounds and / or compositions in more detail, and each reference is incorporated herein by reference in its entirety. Summary of the Invention

[0008] This paper discloses a bioinformatics tool for analyzing biological macromolecules, including peptides, polypeptides, and polypeptides, which complements recently described methods for polypeptide analysis employing nucleic acid-based polypeptide encoding techniques. The disclosed bioinformatics method enables high-throughput processing of nucleic acid-encoded polypeptide data to identify and quantify thousands of peptides and / or polypeptides present in biological samples.

[0009] Recently, methods for high-throughput peptide characterization have been published, such as US 2019 / 0145982 A1, US11513126 B2, and US 2023 / 0136966 A1, which utilize nucleic acid barcoding binding reagents to: cyclically identify specific components of an immobilized peptide and encode the binding history of each peptide in a nucleic acid recording tag after each binding event, thereby generating extended recording tags. Thus, when a portion of the peptide is identified by the nucleic acid barcoding binding reagent, a nucleic acid signal regarding the binding event is generated and recorded in a recording tag associated with the peptide, which can then lead to probabilistic identification of the identified portion. The identified portion (e.g., terminal amino acid residues) is cleaved, creating an opportunity to identify new components of the immobilized peptide (e.g., newly formed terminal amino acid residues). This process can be repeated until sufficient information about the peptide components has been encoded (e.g., 2 to 10 binding / encoding cycles may be sufficient for peptide identification), see also Figure 2 After encoding information about the binding agents that bind to a particular immobilized peptide in an extended record tag associated with the peptide, the extended record tag can be analyzed in parallel, for example, by next-generation sequencing (NGS). Furthermore, information about the peptide's composition is elucidated by decoding information about the binding agents that bind to these peptides during each binding cycle.

[0010] This paper provides a computer-implemented method for analyzing multiple peptides encoded in multiple nucleic acid sequences using a coding assay. The method includes: (a) receiving, at one or more processors, multiple nucleic acid sequences generated from a coding assay, wherein each nucleic acid sequence comprises a series of encoder barcode sequences, and wherein each encoder barcode sequence in the given series corresponds to a conjugate from a set of conjugates that bind to one or more components of the peptide in the multiple peptides; and (b) generating a conjugate identifier string for each nucleic acid sequence in the multiple nucleic acid sequences based on the corresponding series of encoder barcode sequences, thereby generating multiple conjugate identifiers corresponding to the multiple nucleic acid sequences. (c) Using one or more processors, infer the amino acid sequence of a polypeptide from a plurality of binder identifier strings, the inference being based on (i) binding characteristics of binders from a set of binders corresponding to binder identifiers in the binder identifier strings, and (ii) a calculated probability score of the association between one or more binder identifiers in the binder identifier strings and one or more amino acid sequences of the polypeptides in the plurality of polypeptides; and (d) based on the calculated probability score and the inferred amino acid sequence, outputting data relating to at least partial identity and / or quantity of at least one polypeptide in the plurality of polypeptides.

[0011] This paper provides a computer-implemented method for analyzing multiple peptides encoded in multiple nucleic acid sequences using a coding assay, the method comprising: a) receiving, at one or more processors, multiple nucleic acid sequences generated from a coding assay, wherein each of the multiple nucleic acid sequences comprises a series of encoder barcode sequences, and wherein each barcode sequence in the series of encoder barcode sequences corresponds to a conjugate that binds to a component of a peptide in the multiple peptides in the coding assay; b) using one or more processors to assign conjugate identifiers to each encoder barcode sequence in the series of encoder barcode sequences of each of the multiple nucleic acid sequences to generate multiple conjugate identifier strings for the multiple nucleic acid sequences; c) using one or more processors to provide the multiple conjugate identifier strings as input to a trained machine learning model, wherein the trained machine learning model is configured to infer amino acid sequences of peptides in the multiple peptides from the conjugate identifier strings and output data relating to at least partial identity and / or quantity of at least one peptide in the multiple peptides, the inference being based on the binding characteristics of a conjugate corresponding to a conjugate identifier assigned in a given conjugate identifier string.

[0012] This document also provides a computer-implemented method for analyzing multiple peptides encoded in multiple nucleic acid sequences using a coding assay, the method comprising: a) receiving, at one or more processors, multiple nucleic acid sequences generated from a coding assay, wherein each nucleic acid sequence comprises a series of encoder barcode sequences, and wherein each barcode sequence corresponds to a conjugate that binds to a component of a peptide in the multiple peptides of the coding assay; b) using one or more processors to assign conjugate identifiers to each encoder barcode sequence in the series of encoder barcode sequences of each nucleic acid sequence to generate multiple conjugate identifier strings for the multiple nucleic acid sequences; c) for each peptide and / or peptide fragment in the multiple peptides, generating a set of simulated conjugate identifier strings in a computer based on predetermined parameters of the coding assay. The string, wherein predetermined parameters include: i) the probability of correctly or incorrectly assigning one or more conjugate identifiers to at least one component of a given polypeptide or polypeptide fragment based on the binding characteristics of conjugates used in the coding assay; and ii) optionally, for at least one component of a given polypeptide or polypeptide fragment, the probability of successfully cleaving the at least one component after conjugates bind to the at least one component in the coding assay, thereby determining the probability of each simulated conjugate identifier string generated from the given polypeptide or polypeptide fragment; d) for each of the plurality of conjugate identifier strings generated in (b), calculating the probability of a given conjugate identifier string being generated from one or more polypeptides or fragments thereof based on the probability of the simulated conjugate identifier strings determined in (c), thereby determining at least partial identity and / or quantity of at least one polypeptide among a plurality of polypeptides.

[0013] This document also provides a system comprising: one or more processors; and a memory communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to perform the computer-implemented methods described above.

[0014] This document also provides a non-transitory computer-readable storage medium that stores one or more programs, the one or more programs including instructions that, when executed by one or more processors of the system, cause the system to perform the computer-implemented methods described above.

[0015] This overview is not intended to limit the scope of the claimed subject matter. Other features, details, utility, and advantages of the claimed subject matter will become apparent from the detailed description, including those embodiments disclosed in the drawings and appended claims. Attached Figure Description

[0016] Non-limiting embodiments of the present invention will be described by way of example with reference to the accompanying drawings, which are schematic and not intended to be drawn to scale. For illustrative purposes, not every component is labeled in each figure, nor are every component of each embodiment of the invention shown, and the illustrative drawings are not necessary for those skilled in the art to understand the invention.

[0017] Figure 1 An exemplary flowchart of a method for analyzing peptides, illustrating the key steps in sample preparation and analysis.

[0018] Figure 2 An exemplary schematic diagram of a assay (also known as an encoding assay) used for reverse translation to encode a polypeptide (e.g., peptide or polypeptide) sequence into a barcoded nucleic acid sequence.

[0019] Figure 3A An exemplary flowchart of a method for identifying peptides based on the analysis of nucleic acid sequence data generated from peptide coding processes, which utilizes partially selective binders and exhibits a high error rate in interpreting individual amino acid residues.

[0020] Figure 3B Based on the process of encoding peptides (such as...) Figure 2 An exemplary flowchart of a method for analyzing and identifying peptides from nucleic acid sequence data generated (as depicted in the diagram).

[0021] Figures 4A to 4D An exemplary flowchart of a method for identifying peptides based on the analysis of nucleic acid sequence data generated from peptide coding processes, which utilizes partially selective binders and exhibits a high error rate in interpreting individual amino acid residues. Figure 4A Exemplary steps of an embodiment of the method are shown, which include generating a set of simulated combiner identifier strings in a computer based on predetermined parameters of an encoding determination method (see Example 8 for more details), while Figure 4C The general scheme of this embodiment is shown. Figure 4B The exemplary steps of a second embodiment of the method are shown, which include generating multiple peptide reads and a probability score for each peptide read based on the probability of a given peptide read arising from a given binder identifier string (see Example 9 for more details). Figure 4D The general scheme of the second embodiment is shown.

[0022] Figure 5 Exemplary schematic diagrams of computing devices according to one or more embodiments of the present disclosure.

[0023] Figure 6 An exemplary schematic diagram of the structure of a nucleic acid recording tag used to encode peptide and / or polypeptide sequence information.

[0024] Figures 7A to 7B An exemplary schematic diagram of a assay for performing reverse translation to encode peptide and / or polypeptide sequence information into barcoded nucleic acid sequences. Figure 7A : Determination method using only double-stranded linkage. Figure 7B : A method for determining the connection and extension of one chain in a double-stranded region.

[0025] Figure 8 Exemplary barcode-level alignment of nucleic acid sequence reads. The model nucleic acid sequence (SEQ ID NO: 12) was analyzed.

[0026] Figure 9 Generate reads for each analyzed nucleic acid sequence (e.g., ...). Figure 8 An exemplary output of a computer model of at least one binding sub-identifier string (as described in [reference]). The nucleic acid sequence read shown includes the following sequences: SEQ ID NO: 13-SEQ ID NO: 17.

[0027] Figure 10 An exemplary schematic diagram of a proteomics analysis system according to one embodiment as described herein.

[0028] Figure 11 An exemplary diagram illustrating potential errors that can occur when encoding polypeptide sequence information as barcoded nucleic acid sequences.

[0029] Figure 12 Exemplary false discovery rate (FDR) data as a function of peptide concentration in samples analyzed using publicly available assays and bioinformatics techniques.

[0030] Figure 13 An exemplary schematic diagram illustrating the differences between segment-level analysis and batch-level analysis.

[0031] Figure 14 An exemplary schematic diagram illustrating the use of the Expectation-Maximization (EM) algorithm to refine the prediction of peptide abundance in a sample from the output of a trained machine learning model.

[0032] Figure 15 Exemplary simulated false discovery rate (FDR) data as a function of the relative abundance of peptides in samples analyzed using publicly available assays and bioinformatics techniques.

[0033] Figures 16A to 16G Exemplary heatmap data is shown, illustrating the coding efficiency of seven engineered binders having the sequences described in SEQ ID NO:1 to SEQ ID NO:7, used as a group of binders in a multiple coding assay of at least 289 immobilized peptides. The more intense the white of the unit representing a specific P1-P2 combination, the higher the coding efficiency.

[0034] Figure 17 Exemplary heatmap data is shown, illustrating the coding efficiency of independent binders having the sequence described in SEQ ID NO:7, used in multiple coding assays at different concentrations (15 nM, 31 nM, 63 nM, 125 nM, 250 nM, and 500 nM). The more intense the white of the unit representing a specific P1-P2 combination, the higher the coding efficiency.

[0035] Figures 18A to 18B The theoretical modeling and experimental data are presented, demonstrating the dependence of the coding yield of a single binder on the binder concentration in the coding assay. Figure 18A An exemplary modeling of the coding yield of independent binders, which depends on the binder concentration [A], is shown. Figure 18B Exemplary experimental data showing concentration-dependent encoded yields of two independent engineered binders are presented.

[0036] Figures 19A to 19B An exemplary method is shown for estimating the NTF efficiency of an N-terminal modifier that reacts with NTAA residues of a peptide. Figure 19A An exemplary scheme is shown that utilizes a biotinylated N-terminal modifier and detects successful NTAA installation by binding to streptavidin and encoding binding information. Figure 19B Exemplary experimental heatmap data show the coding efficiency of specific P1-P2 residues of the immobilized peptide.

[0037] Figure 20 An exemplary calculation of the modified NTAA cleavage efficiency (NTC efficiency) of an immobilized peptide group is shown, which is performed using experimental heatmap data and a specific cleavage enzyme acting on an array of peptides with different P1-P2 residues.

[0038] Figures 21A to 21B An exemplary simulation process is shown to generate the probability distribution of the simulated binder identifier string, as described in Example 8. A1A2A3… represents the sequence of the peptide being analyzed. Figure 21A This illustrates the potential consequences of cleavage of terminal amino acid residues after the first and second coding cycles. Figure 21B The results of the assignment of binder identifiers for a given terminal amino acid residue within the coding cycle are shown. For residue A1, BBI, BI1, and BI2 were assigned (based on simulated binding of binding agents 1 and 2). For residue A2, BBI, BI3, and BI4 were assigned (based on simulated binding of binding agents 3 and 4).

[0039] Figures 22A to 22B An exemplary calculation of the probability score of a peptide read generated using a model system for two specific binder identifier strings is shown.

[0040] Figures 23A to 23B Another example is shown of calculating the probability of a peptide read based on a given binder identifier string “LVNDVZDF” (SEQ ID NO: 18). The generated peptide read contains the amino acid sequence as described in SEQ ID NO: 19 to SEQ ID NO: 27.

[0041] Figure 24 An exemplary generation of a large number of custom peptide-polynucleotide conjugates immobilized on a solid support is shown, which can be used for large-scale coding assays, and the data is used to train machine learning models (see Example 10). Detailed Implementation

[0042] Highly parallel characterization and identification of peptides remains a challenge. In proteomics, a goal is to identify and quantify large numbers of peptides in a sample, and achieving this in a high-throughput manner is a daunting task. Assays such as immunoassays and mass spectrometry-based methods have been used, but these methods are limited at both the sample and analyte levels, with limited sensitivity and dynamic range, and are subject to cross-reactivity and background signal. Multiplexing readouts from a set of affinities to a set of homologous peptides (e.g., using affinities with detectable labels) remains challenging.

[0043] Recently, a novel form of high-throughput peptide sequencing assay (referred to as NGPS (Next-Generation Peptide Sequencing) assay, ProteoCode® assay, or coding assay) was reported in published U.S. patent applications US 2019 / 0145982 A1 and US2023 / 0136966 A1, see also Figure 2 and Figures 7A to 7B During the exemplary assay, immobilized peptides, each associated with a nucleic acid recording tag, are contacted with a set of binding reagents (or binders), each capable of binding to a different set of modified N-terminal amino acid (NTAA) residues of the peptide, wherein each binding reagent includes a nucleic acid coding tag having identification information about the binding reagent. During each binding cycle, the coding tag and the recording tag are placed close enough to interact, and information about the binding reagent bound to the peptide in the current cycle is transferred between the coding tag and the recording tag, for example, from the coding tag to the recording tag, thereby generating an extended recording tag associated with the peptide. After the transfer is complete, the modified NTAA residues of the peptide are cleaved with a specific reagent (see, for example, US11427814 B2 or US 2022 / 0227889 A1), thereby generating new NTAA residues of the peptide, which can be measured in a similar manner in the next binding cycle using a new set of binding reagents.

[0044] After at least one consecutive binding cycle, a library of nucleic acid-encoded DNA representing the binding history of each peptide is generated, where information about each binding event is encoded in an extended recording tag. After analyzing the extended recording tags (typically via nucleic acid sequencing methods), information about the binding reagent (bindonic) that bound to the peptide at each cycle can be decoded, providing information about the components of the peptide bound by the binding reagent (e.g., NTAA residues). Therefore, the disclosed variants of the ProteoCode® assay (collectively referred to throughout this disclosure as the “encoded assay”) represent an unconventional approach to the parallel characterization, identification, or quantification of large numbers of peptides and provide a means of sequentially identifying independent amino acid residues of peptides with a certain probability.

[0045] In some embodiments, coding assays generate a nucleic acid-encoded library representation of the binding history of each of a variety of target peptides (i.e., peptide analytes). This nucleic acid-encoded library can be amplified and analyzed using high-throughput next-generation digital sequencing methods, enabling the analysis of millions to billions of molecules per run. Creating nucleic acid-encoded libraries of binding information is also useful because it allows for enrichment, reduction, and normalization using DNA-based techniques that leverage hybridization. These DNA-based methods are easily and rapidly scaled and customized, and are more cost-effective than methods that can be used to directly manipulate other types of large molecule libraries, such as peptide libraries. Therefore, nucleic acid-encoded libraries of binding information can be processed prior to sequencing using one or more techniques to enrich and / or reduce and / or normalize the sequence representation. This enables the extraction of the most interesting information from extremely large libraries, where the abundance of individual members can initially vary over many orders of magnitude.

[0046] In some embodiments, analysis of the composition of the extended recording tag (e.g., by determining the incorporated nucleic acid moieties using nanopore sequencing or by sequencing the entire extended recording tag) is used to obtain information about one or more binders specifically binding to the analyzed peptide and the order in which these binders bind. The binding characteristics and binding order of the binders provide identification information about specific components of the analyzed peptide bound by the binders (e.g., modified NTAA residues) and about the overall sequence of the analyzed peptide. As a result of the above-described encoding cycle, the structural information of the immobilized peptide is encoded as a nucleic acid sequence. Thus, the identity of the immobilized peptide can be decoded from the sequence of the corresponding associated recording tag by calculating the probability of occurrence of specific types of amino acid residues at corresponding positions in the amino acid sequence of the immobilized peptide. The binders in the method do not need to be strictly selective and can identify, for example, the structural / functional categories of NTAA residues, such as negatively charged residues, positively charged residues, small hydrophobic residues, aromatic residues, etc., or identify other types of NTAA residues. This is because several binders can be used simultaneously to decode a single type of NTAA residue.

[0047] In a preferred embodiment of the coding assay, after several cycles of binding / transfer, each immobilized peptide is back-translated into a series of unique nucleic acid barcodes on a corresponding recording tag associated with the immobilized peptide. During the analysis step, the sequence of the extended recording tag can be analyzed to extract the aforementioned nucleic acid barcodes corresponding to each coding cycle. Then, to associate the extracted nucleic acid barcodes with corresponding amino acid residues, an artificial intelligence (AI) model can be applied to calculate the probability of occurrence of specific types of amino acid residues at corresponding positions in the amino acid sequence of the analyzed peptide. In some embodiments, the AI ​​model can be pre-trained using multiple known peptide sequences used to generate nucleic acid coding data associated with the recording tags. Modeling the coding of multiple known peptides using a known set of binders enables the AI ​​model to faithfully predict amino acid residues based on the provided encoder barcode sequences.

[0048] Existing challenges in high-throughput peptide identification using coding assays include the absence of conjugates with strong selectivity for independent terminal amino acid residues of the peptide analyte, leading to high error rates in independent amino acid interpretation during analysis. Most conjugates exhibit “diffusion” specificity (they can interact with a number of structurally similar NTAAs with a certain probability). Furthermore, most conjugates exhibit “P1P2 specificity,” meaning their affinity for a specific terminal amino acid residue (i.e., the P1 residue) depends on the penultimate terminal amino acid of the peptide (i.e., the P2 residue). Therefore, conjugates tend to bind better to peptides with certain combinations of P1 and P2 residues, further complicating the analysis and obscuring information about the encoded NTAAs. Other sources of error in the coding data include: i) the absence of a signal in a specific binding cycle due to failed coding events (failure to transfer identification information to the extended RT); and ii) incorrect signals generated due to nonspecific or off-target binding. Nonspecific binding occurs when conjugates bind nonspecifically to other components of the support or system. Off-target binding occurs when conjugates bind to non-homologous components of the peptide (e.g., to NTAAs they should not bind to). Such non-specific or off-target binding can occasionally generate misleading coding signals. All the factors discussed contribute to the significant variability in coding performance and form the basis for errors in the identification of specific residues and the overall identity of the peptide.

[0049] Due to the potential for errors introduced by coding assays, computer models (e.g., including algorithms based on AI or machine learning running on one or more processors) can be used to identify potential sources of error, predict the likelihood of generating multiple binder identifier “signature” reads from a known peptide, and then calculate the probability of identifying a specific peptide based on existing binder identifier string data. Due to the inherent complexity of assays (as described above), multiple proteomics “signature” reads (e.g., tens, hundreds, or even thousands; also referred to herein as “peptide reads”) can be generated from a single peptide as a coding signal, and the computer models discussed herein are capable of matching the generated proteomics “signature” reads back to a single peptide.

[0050] Exemplary bioinformatics analysis workflow:

[0051] As described above, the output of a coding assay for analyzing multiple peptides includes a library of nucleic acid sequences, wherein each nucleic acid sequence in the library corresponds to an extended recording tag associated with a peptide derived from one of the analyzed peptides. Therefore, each nucleic acid sequence in the library contains a series of encoder barcode sequences, wherein each encoder barcode sequence in the series indicates a binder to the analyzed peptide derived from one of the analyzed peptides. In some embodiments, each nucleic acid sequence in the library further includes at least one auxiliary sequence, wherein the at least one auxiliary sequence includes one or more sample barcode sequences, one or more bead barcode sequences, one or more unique molecular identifier (UMI) sequences, one or more spacer sequences, sequencing primer sequences, and / or their complements or combinations thereof.

[0052] In some embodiments, a library of nucleic acid sequences is obtained from an NGS device and received at one or more processors for further analysis. In these embodiments, initial NGS data processing can be performed using publicly available tools. Exemplary, non-limiting NGS data processing tools are provided below.

[0053] a) Create FASTQ format files from raw sequencing files using bclconvert or bcl2fastq. FASTQ is a text-based sequence file format generated from binary base read (BCL) files (raw data files generated by Illumina sequencers), storing both raw sequence data and quality scores. Alternative processing tools can be used. For example, Illumina BCL Convert software is a Linux application that converts BCL files generated by Illumina sequencing systems into FASTQ format files.

[0054] b) Obtain quality control statistics using FastQC / MultiQC tools. Quality control of nucleic acid samples from sequencers is crucial for further production mapping. FastQC aims to provide a simple way to perform quality control checks on raw sequence data from high-throughput sequencers. It offers a modular set of analyses that can be used to quickly check the data for any issues that the user should be aware of before proceeding with any further data analysis. MultiQC is a program that creates summaries of all samples for several different types of QC measures. MultiQC aggregates results from bioinformatics analyses across many samples into a single report. It retrieves analysis logs in a given directory and compiles HTML reports. MultiQC automatically locates and summarizes output files from supported tools.

[0055] c) Merging paired-end reads with BBMerge. BBMerge is designed to merge two overlapping paired reads into a single read (Bushnell B, Rood J, Singer E. BBMerge - Accurate paired shotgun readmerging via overlap. PLoS One. 2017 Oct 26;12(10):e0185056). Merging paired-end reads improves a variety of subsequent bioinformatics processes. BBMerge also provides the ability to merge non-overlapping shotgun read pairs by assembling unsequencing gaps between reads using k-mer frequency information, thereby achieving significantly higher merging rates while maintaining or improving accuracy.

[0056] Alternative processing tools may also be used. Following initial processing, a library of multiple nucleic acid sequences is generated from the nucleic acid sequences received from the NGS device. In some embodiments, the multiple nucleic acid sequences exist in a text-based sequencing data file format (such as FASTQ), which stores both the raw sequence data and quality scores. This format can be used as input for various secondary data analysis schemes.

[0057] In the next step, multiple identifier strings are generated for multiple nucleic acid sequences. This generation is performed by assigning a binding sub-identifier to each of a series of encoder barcode sequences using one or more processors within each of the multiple nucleic acid sequences, and optionally assigning an auxiliary identifier to at least one of one or more auxiliary sequences. In some embodiments, nucleic acid sequences in FASTQ format provided from previous processing steps are aligned and the binding sub-identifier and auxiliary identifier are assigned. Exemplary assignments are as follows: Figure 8 and Figure 9As shown in the illustration. In these embodiments, the designed extended RT construct includes a sequence of barcodes, spacers, and other elements present in each extended RT and thus in each of the plurality of nucleic acid sequences generated after initial processing of the raw sequencing data (see, for example, SEQ ID NO:12 to SEQ ID NO:17). Exemplary barcodes designed to be present in a specific order in the extended RT include: BBC barcodes, RT barcodes, CT barcodes, spacer sequences, LM sequences, and CBC barcodes attached during the final end-capping step, which represent the end of a sequence read. In some embodiments, some of these sequences or barcodes are not present in the extended RT. In some embodiments, other sequences or barcodes may be present in the extended RT. During the step, a binding identifier is assigned to each CT barcode present in each nucleic acid sequence of the analyzed nucleic acid sequence, and an auxiliary identifier is assigned to at least one of the auxiliary sequences in each nucleic acid sequence of the analyzed nucleic acid sequence. In some embodiments, for assignment purposes, all other barcodes and identification sequences besides the CT barcode are labeled as “auxiliary sequences.” The key difference between CT barcodes and auxiliary sequences is that CT barcodes include identification information about the binding reagent that binds to a particular polypeptide during one or more binding cycles. Therefore, given a defined binding characteristic of the binding reagent, analysis of the CT barcode corresponding to a particular analyzed polypeptide provides actual structural information about the specific analyzed polypeptide (e.g., the identity of specific amino acids within the polypeptide sequence). Conversely, auxiliary sequences provide additional information about the analyzed polypeptide, such as which sample the polypeptide originated from, which compartment or fraction the polypeptide originated from, or what type of polypeptide the polypeptide originated from (e.g., when the polypeptide was formed during polypeptide fragmentation). Furthermore, auxiliary sequences may correspond to one or more optional elements of the extended RT construct, including spacer sequences (typically present between CT barcodes), sequences to be used for amplification in extended RT, or sequences representing the start or end of a sequence read. In some embodiments, auxiliary identifiers are assigned to all auxiliary sequences in each nucleic acid sequence of the analyzed nucleic acid sequence. Typically, each auxiliary sequence has a specific purpose in peptide identification assays; for example, BBC / RT barcodes identify the sample from which the analyzed peptide originates, and spacer sequences are required for coding to occur (in some embodiments).

[0058] In some embodiments, the computer model used to assign binder identifiers to each encoder barcode sequence in the encoder barcode sequence and optionally to one or more auxiliary identifiers to at least one of the auxiliary sequences is Talon-LUT, which uses a lookup table to quickly assign tags (i.e., identifiers) to individual segments of the analyzed nucleic acid sequence. It uses a "syntax" to know which "word" (i.e., the type of barcode or auxiliary sequence) is next in the designed extended RT construct. It performs error correction by enumerating all common types of errors for each known barcode or auxiliary sequence. In some embodiments, common types of errors include DNA sequencing errors such as single nucleotide polymorphisms (SNPs) and insertions / deletions. An example of such error correction is as follows. For the barcode ACCCCGT, the enumerated errors can be assigned by the computer model with a certain probability (such as the enumerated errors CCCCCGT, GCCCCGT, AACCCGT, etc.). If the computer model encounters a read containing CCCCCGT, it is not an exact match with the barcode, but it is an exact match with one of the enumerated errors. Therefore, the computer model matches it to the enumerated incorrect sequence and also to its associated exact barcode.

[0059] In some embodiments, the computer model used to assign a binding identifier to each encoder barcode sequence in the encoder barcode sequence and optionally to one or more auxiliary identifiers to at least one of the auxiliary sequences is a Talon-HMM, which models each sequencing read using a Hidden Markov Model (HMM) (or another similar probabilistic model). In some embodiments, the Baum-Welch algorithm is used to find the most likely sequence for the tag of each read. The Baum-Welch algorithm is a special case of the Expectation-Maximization (EM) algorithm. Its purpose is to tune the parameters of the HMM, namely the state transition matrix A, the emission matrix B, and the initial state distribution π0, so that the model most closely resembles the observed data. Although Talon-HMM is slower than Talon-LUT, it provides higher accuracy for assignment.

[0060] In some embodiments, additional tools are used to generate sample-level statistics for multiple nucleic acid sequences.

[0061] In the next step, one or more processors provide multiple identifier strings generated in the previous steps as input to the computer model. In some embodiments, the multiple identifier strings may be further processed or transformed before being provided to the computer model as input. In some embodiments, the multiple identifier strings are converted into multiple peptide reads based on determined binding characteristics of the binders, which correspond to binder identifiers assigned in a given binder identifier string. In some embodiments, the computer model reads the multiple identifier strings generated by Talon-LUT or Talon-HMM, identifies the binder identifiers within the multiple identifier strings, and generates “peptide reads” corresponding to a series of possible encoded events in the coding assay. In a preferred embodiment, the multiple peptide reads are generated from a given identifier string, depending on the complexity of the binding characteristics of the binders that bind to the analyzed peptide during the coding assay and taking into account the probabilistic nature of other steps in the coding assay (efficiency of NTF, NTE, and NTC steps; see also Examples 7 and 9 below). In some embodiments, the multiple identifier strings are not converted into multiple peptide reads before being used as input to the computer model.

[0062] In some embodiments, the multiple binder identifier strings generated in the previous steps are evaluated against a variety of peptides that may be present in the analyzed sample. In some embodiments, the multiple binder identifier strings are provided as input to a trained machine learning (ML) model, wherein the ML model is configured to assign the binder identifier strings to the amino acid sequences of peptides (which may be present in all peptides in the analyzed sample) and output at least partial identity and / or quantity of at least one peptide among the multiple peptides. Two different workflows that can be used for this step are described below.

[0063] 1) Batch analysis algorithm.

[0064] Batch analysis algorithms find unique characteristic spectra associated with the quantity of a specific peptide in a sample. These unique spectra are identified from training data. A selectable statistical threshold ensures that the characteristic spectra read patterns frequently appear from the target peptide but rarely from the background. The output of the batch analysis algorithm is quantitative information at the sample level (e.g., "peptide X is present in this sample at every million Y peptides") and may include probability assignments of reads to individual peptides. The algorithm's input includes two types of training data: a) training data from pure peptide samples of interest (e.g., recombinant peptides of interest that can generate identifiable specific characteristic spectra reads) – "foreground" samples; and b) training data from "background" samples, such as plasma samples. Batch analysis algorithms require training data from both the foreground and background to analyze realistic peptide samples. Using this training data, it selects unique characteristic spectra that are constant in the background and simultaneously uniquely identifies the foreground (the specific peptide of interest). For each sample of interest, it uses these identified unique characteristic spectra to predict the quantity of each foreground peptide. The batch analysis algorithm does this by correcting the peptide observations (based on the unique characteristic spectra) with a correction factor calculated from the training data. The correction can be calculated based on the known true positive rate and false positive rate in the foreground sample and background sample, respectively.

[0065] Next, the batch analysis algorithm refines the quantification of the predictions using Expectation-Maximization (EM). EM is an iterative algorithm for refining model predictions. EM iterations alternate between Expectation (E) and Maximization (M) steps, where the Expectation step creates a function of the expectation of the likelihood evaluated using the current estimates of the parameters, and the Maximization step computes the parameters that maximize the expected likelihood found in the E step. These estimates of the parameters are then used to determine the distribution of the latent variables in the next E step. In some embodiments, the following two steps are repeated in the EM until convergence: 1) Expectation: Given the current model parameters (current estimates of what peptides are in the sample), find the optimal assignment (in this case, the input binder identifier string) for each data point (in this case, the peptide label). 2) Maximization: Given the current assignment, find the optimal estimate of the model parameters (e.g., the current CT barcode assignment).

[0066] 2) Read-level analysis. The output of read-level analysis is a mapping for each read (e.g., "Read X is from peptide Y"), and quantitative information at the sample level (e.g., "Peptide X is present in this sample at one million Y peptides"). Read-level analysis can be performed in several different ways. A more detailed description is provided in Example 8 below.

[0067] (2A) A Salmon-inspired "top-down" algorithm (Patro R, et al., Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods. 2017 Apr;14(4):417-419), but with different details. It uses an HMM-like probabilistic model trained on empirically determined parameters to construct an "index" that maps short read segments (called "k-mers") to probabilities. For each binder identifier string, it breaks it down into k-mers, determines the probability of each k-mer, and then determines whether the binder identifier string can be assigned to a specific peptide based on the determined probabilities of a set of k-mer segments. The k-mers are constructed from a specific binder identifier string as follows: for the exemplary read "A1A2A3A4A5A6A7", all k-mers (pentamers) of length 5 would be A1A2A3A4A5, A2A3A4A5A6, and A3A4A5A6A7. The assignment of a combination sub-identifier string to a specific polypeptide can be based on the product of the independent probabilities of all k-mers in the group, or on some other functional index related to the k-mer probability.

[0068] In some embodiments, the parameters of the computer model are categorized as "emission probability" (the probability of seeing barcode X given an underlying amino acid of Y; essentially, the encoding probability) and "transition probability" (the probability of continuing to the next amino acid after encoding; i.e., the likelihood that a lyase will cleave NTAA residues after encoding to expose new NTAA residues of the analyzed polypeptide). In some embodiments, a "top-down" algorithm learns the parameters (emission and transition probabilities) of the encoding system from training data and then applies that knowledge to real data.

[0069] (2B) Deep learning-based segment mapping.

[0070] For this algorithm, the same empirically determined parameters as the previously described "top-down" algorithm are used. The computer model generates multiple simulated encoded data sets from the model peptide. The computer model creates binder identifier strings based on a given input distribution of the peptide using probabilities (including the emission and transition probabilities described above). In some embodiments, a set of simulated binder identifier strings includes at least 100,000, at least 500,000, or more simulated binder identifier strings. Therefore, the generated simulated data is used to train a deep learning model to map from the generated raw reads to probabilistic peptide assignments.

[0071] Next, real reads from actual coding assays (rather than simulated reads) are used to further train the deep learning model to learn aspects of the erroneous model not captured in the empirical assay parameters. The advantage of using simulated data first is that several orders of magnitude more training data can be created through simulation rather than from actual assays. The deep learning model is used to assign each binder identifier string in the sample to a peptide that may be present in the sample in a fractional manner. As an example of fractional assignment: "For read X, the deep learning model predicts that it has a 30% probability of coming from peptide A and a 70% probability of coming from peptide B."

[0072] Next, the Expectation Maximization (EM) algorithm is used to refine the predicted read allocation and quantification in a similar manner to that described above for the batch analysis algorithm.

[0073] Bioinformatics tools for analyzing biological macromolecules (e.g., peptides, polypeptides, and peptides as described herein) complement recently described methods of polypeptide analysis employing nucleic acid-based polypeptide encoding technologies (i.e., next-generation polypeptide assays (NGPA)). See, for example, U.S. Patent Application Publications US 2019 / 0145982 and US 2023 / 0136966 A1, each incorporated herein by reference in its entirety. The disclosed bioinformatics methods enable high-throughput processing of nucleic acid-encoded polypeptide data to identify and quantify thousands of peptides and / or polypeptides present in biological samples using a limited set of binders (e.g., 5 to 20 binders with imperfect selectivity for the terminal amino acid residues of peptides).

[0074] Figure 1 Non-limiting examples of flowcharts for identifying and / or quantifying peptides, polypeptides, and peptides extracted from samples are provided, including steps of coded assays (e.g., 106, 108, and 110).

[0075] exist Figure 1 In step 102, peptides, polypeptides, and polypeptides are extracted from the sample and fragmented. In some cases, the sample may be, for example, a tissue sample or a cell extract sample. In some cases, the sample may be a cell extract sample, and the cell extract sample may be a mammalian cell extract, plant cell extract, fungal cell extract, or bacterial cell extract.

[0076] Peptides, polypeptides, and peptides can be extracted from samples using any of a variety of techniques known to those skilled in the art, wherein the chosen technique may depend on the sample type and whether the extracted polypeptides in the lysate are to be analyzed in a non-denatured or denatured state. For NGPA assays, native or denatured polypeptides can be utilized. Examples of steps that can be combined to perform peptide and polypeptide extraction include, but are not limited to: homogenization of tissue samples (e.g., by grinding), lysis of cell suspensions (e.g., by sonication or application of shear force), polypeptide precipitation using acetone (for preparing denatured lysates), dissolution and / or resuspension in a suitable polypeptide extraction buffer, immunoprecipitation, filtration, centrifugation, biomagnetic separation, and / or affinity chromatography.

[0077] Peptides and peptides extracted from samples can be fragmented using any of a variety of techniques known to those skilled in the art. Examples include, but are not limited to, treatment with specific proteases or endopeptidases (e.g., TEV proteases, which are specific for the ENLYFQ\S common amino acid sequence), treatment with non-specific proteases or endopeptidases (e.g., peptidase K), and / or treatment with chemical reagents that cleave peptide bonds (e.g., cyanogen bromide (CNBr), hydroxylamine, hydrazine, etc.).

[0078] In some cases, peptide and / or peptide fragmentation can be performed before attaching a DNA tag (e.g., a DNA recording tag as described below). In other cases, peptide and / or peptide fragmentation can be performed after attaching a DNA tag (e.g., a DNA recording tag as described below).

[0079] In some cases, peptides, polypeptides, and / or peptides may undergo one or more additional fractionation steps before attachment to a solid support. For example, peptides, polypeptides, and / or peptides may be separated (or enriched) by one or more properties such as cellular location, molecular weight, hydrophobicity, or isoelectric point. Alternatively or additionally, one or more polypeptide enrichment steps may be used to select specific polypeptides or peptides (see, for example, Whiteaker, et al. (2007), Anal. Biochem. 362:44-54) or to select specific post-translational modifications (see, for example, Huang, et al. (2014), J. Chromatogr. A 1372:1-17). Alternatively, specific classes of polypeptides (such as immunoglobulins or immunoglobulin (Ig) isoforms, such as IgG) may be affinity enriched or selected for analysis. Overly abundant polypeptides may also be reduced from the sample using standard immunoaffinity methods. The depletion of abundant polypeptides can be useful, for example, for plasma samples in which more than 80% of the polypeptide composition is albumin and immunoglobulin. Several commercial products can be used to consume excessively abundant peptides in plasma samples (see, for example, PROTIA and PROT20 (Sigma Aldrich, St. Louis, MO)).

[0080] exist Figure 1 In step 104, multiple peptides, polypeptides, or polypeptide fragments are anchored together with corresponding recording tags to multiple solid supports (e.g., beads or other substrates).

[0081] In some cases, multiple peptides, polypeptides, or polypeptide fragments may include at least 100, 500, 1,000, 1,500, 2,000, 2,500, 5,000, 7,500, or 10,000 different peptides, polypeptides, or polypeptides.

[0082] In some cases, peptides, polypeptides, or polypeptide fragments may be co-located or co-labeled with one or more recording tags. In some cases, the recording tag may include one or more universal initiation sites (or their complements), one or more sequencing primer sites (or their complements), one or more barcodes (e.g., sample barcodes, partition barcodes, compartment barcodes, fractionation barcodes, bead barcodes) or their complements, one or more optional unique molecular identifier (UMI) sequences (or their complements), one or more spacer sequences (for information transfer from the coding tag to the extended recording tag, as described below) (or their complements), one or more auxiliary sequences (or their complements), or any combination thereof. The one or more spacer sequences may be constant across all binding cycles, may be binder-specific, or may be binding cycle-number-specific. The one or more auxiliary sequences may include, for example, one or more sample barcode sequences for identifying the sample origin of a specific polypeptide or polypeptide analyte in the extended recording tag sequence, or may include “milestone” barcodes to determine whether coding is complete.

[0083] In some cases, peptides, polypeptides, or polypeptide fragments may be covalently anchored to (or immobilized on) a solid support. In other cases, peptides, polypeptides, or polypeptide fragments may be non-covalently anchored to (or immobilized on) a solid support. Examples of suitable covalent and / or non-covalent techniques for immobilizing peptides, polypeptides, or polypeptide fragments are described in U.S. Patent Application Publication No. 2019 / 0145982, which is incorporated herein by reference in its entirety.

[0084] In some cases, the solid support can be, for example, nanoparticles, microspheres, beads (e.g., polystyrene beads, polymer beads, agarose beads, acrylamide beads, glass beads, controlled-pore glass beads), solid beads, porous beads, paramagnetic beads, porous substrates, arrays, glass surfaces, silicon surfaces, plastic or polymer surfaces, filters, membranes, silicon wafers, flow cells or microfluidic chips, biochips including signal transduction electronic devices, microtiter wells or ELISA plates, etc.

[0085] exist Figure 1 In step 106, the peptide, polypeptide, or polypeptide fragment is encoded into a nucleic acid sequence through a cyclic process comprising: contacting the immobilized peptide, polypeptide, or polypeptide fragment with a set of binders (each including an attached coding tag), and extending the corresponding recording tag using a "reverse translation" process to transfer information from the coding tag attached to each binder to the extended recording tag (see also...). Figure 2 ).

[0086] The conjugate may include, for example, an antibody or fragment thereof, anticalin, an N-recognition protein polypeptide (e.g., an ATP-dependent Clp protease adaptor polypeptide (ClpS)), aptamers, etc., and variants or homologs thereof, and may further include a coding tag that interacts with the immobilized peptide, polypeptide, or polypeptide fragment, which is co-located and / or co-labeled with one or more recording tags. In some embodiments, the conjugate includes NTAA conjugates as disclosed in US 10852305 B2, US2022 / 0283175 A1, and US 2023 / 0220589 A1 (incorporated herein by reference).

[0087] In some cases, the bound immobilized peptide, polypeptide, or polypeptide component may include a peptide, polypeptide, or a portion thereof obtained by fragmentation of a polypeptide molecule. In some cases, the bound immobilized peptide, polypeptide, or polypeptide component may include one or more amino acid residues (e.g., 1, 2, or 3 amino acid residues). In some cases, the bound immobilized peptide, polypeptide, or polypeptide component may include post-translational modifications. In some cases, the bound immobilized peptide, polypeptide, or polypeptide component may include modified amino acid residues, such as modified terminal amino acid residues (NTAA or C-terminal amino acid (CTAA) residues).

[0088] In some cases, the coding tag may include: an encoder sequence that provides identification information of the binders used in a given reverse translation cycle (i.e., coding cycle), an optional UMI, and a spacer sequence that hybridizes with a complementary spacer sequence on the recording tag, thereby facilitating the transfer of coding tag information to the recording tag (e.g., by primer extension, also referred to herein as polymerase extension).

[0089] like Figure 1 As shown, the cyclic process of contacting an immobilized peptide, polypeptide, or polypeptide fragment with a binding phase including a coding tag and performing reverse translation to transfer information about the identity of a portion of the immobilized peptide, polypeptide, or polypeptide fragment to a corresponding extended recording tag can be repeated multiple times, for example, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, or more than 100 cycles.

[0090] After completing a series of cycles for binding and encoding reactions, the extended recording tag can be converted into an amplifiable library using, for example, a capping cycle step, whereby, for example, a capping sequence comprising a universal primer sequence P1' linked to a universal primer sequence P2 and a spacer sequence Sp' is initially annealed with the extended recording tag via complementary P1 and P2' sequences to bring the end cap close to the extended recording tag. The complementary Sp and Sp' sequences in the extended recording tag and end cap are annealed, and primer extension adds a second universal primer sequence (P2) to the extended recording tag.

[0091] exist Figure 1 In step 108, extended recording tags are merged and used to prepare a DNA sequencing library representing multiple immobilized peptides, containing information about the binding history of the peptides. The term "binding history" of a peptide refers to the identity and sequence of binding reagents that bind to and encode one or more components of the peptide in the coding assay.

[0092] In some cases, a library may include at least 5,000, 10,000, 20,000, 30,000, 40,000, 50,000, 60,000, 70,000, 80,000, 90,000, 100,000, 200,000, 400,000, 600,000, 800,000, 1M, or more than 1M nucleic acid sequences. The library can be appropriately modified using known techniques to enable sequencing on any next-generation sequencing (NGS) platform.

[0093] exist Figure 1 In step 110, the DNA library is sequenced using, for example, next-generation sequencing (NGS) technology to generate multiple sequence reads.

[0094] exist Figure 1 In step 112, multiple sequence reads are mapped to multiple binder identifier strings, which are then used to identify and / or quantify peptides extracted from the sample.

[0095] In some cases, multiple concatenation identifier strings include at least 100,000, 250,000, 500,000, 750,000, 1M, 5M, 10M, 25M, 50M, 75M, 100M, 200M, 300M, 400M, or 500M concatenation identifier strings.

[0096] Figure 2An exemplary peptide sequencing assay using an N-terminal amino acid (NTAA)-specific binding reagent is described. (202) Each peptide molecule is associated with a recording tag (RT) (e.g., DNA) and attached to a bead with a low peptide / RT pair (e.g., peptide / RT conjugate) density (sparseness that allows only intra-peptide / RT pair information transfer). For example, the peptide and its RT can form constructs, and in the case of peptides covalently attached to RTs to form single molecules, the sparsity of constructs on the bead allows only intramolecular information transfer, not information transfer between adjacent constructs on the bead. The N-terminal amino acid (NTAA) residues of the peptide are labeled with an N-terminal modification (NTM). (204) Next, the immobilized and labeled peptide is contacted with a binding reagent specific for the labeled NTAA (a labeled F-specific binding reagent is shown). Each binding reagent includes an encoding tag (CT) (e.g., DNA) that includes identification information about the conjugate in the binding reagent. Following binding and washing, the encoding tag identification information is enzymatically transferred to the recording tag (via extension and / or ligation, such as primer extension and subsequent ligation), generating an extended RT. (206) The labeled NTAAs are removed, for example, by using a mild Edman-like elimination chemical or by a lyase. Cycle 1-2-3 is repeated n times. After n cycles, an extended RT representing n amino acids of the peptide sequence is formed and can be sequenced by NGS. (208) A representative structure of the extended RT after 7 cycles is shown. Figure 2 The exemplary assay described herein may be referred to as next-generation peptide sequencing (NGPS).

[0097] Figure 3A Procedure 300A (corresponding to) is provided for processing barcoded nucleic acid sequences to identify (or partially identify, such as identifying with a certain probability) the polypeptide represented therein. Figure 1The flowchart for step 112) is a non-limiting example. Process 300A may be performed, for example, using one or more electronic devices implementing the software platform. In some instances, process 300A is performed using a client-server system, and the blocks of process 300A are divided in any way between server and client devices. In other instances, the blocks of process 300A are divided between server and multiple client devices. Therefore, although parts of process 300A are described herein as being performed by a specific device of a client-server system, it should be understood that process 300A is not limited thereto. In other instances, process 300A is performed using only client devices or only multiple client devices. In process 300A, some blocks are optionally combined, the order of some blocks is optionally changed, and some blocks are optionally omitted. In some instances, additional steps may be combined with process 300A. Therefore, the operations shown (and described in more detail below) are exemplary in nature and should therefore not be considered limiting.

[0098] exist Figure 3A At step 302A, raw NGS data of multiple peptides or fragments thereof (e.g., peptides or polypeptides extracted from a sample) encoded in multiple nucleic acid sequences (i.e., extended record tag sequences) are received (e.g., by one or more processors of a system configured to implement the disclosed method). The raw NGS data is processed to create and output a sequence read file (e.g., a FASTQ file) comprising nucleic acid sequence reads corresponding to the multiple nucleic acid sequences (corresponding to the extended record tag sequences), wherein each of the multiple nucleic acid sequences contains a series of encoder barcode sequences and optionally one or more auxiliary sequences, and wherein each barcode sequence in the series of encoder barcode sequences corresponds to a binder of a component of a given peptide that binds to the multiple peptides.

[0099] FASTQ files can be created using tools such as BCL Convert (Illumina, Inc., San Diego, California) or bcl2fastq (Illumina, Inc., San Diego, California). Quality control statistics can be generated using tools such as FastQC (BaseSpace Labs, Illumina, Inc., San Diego, California) or MultiQC (Seqera Labs, Barcelona, ​​Spain). Overlapping paired-end reads can be merged using tools such as BBMerge (JointGenome Institute, Lawrence Berkeley National Lab, Berkeley, California).

[0100] As described above, the conjugate may include, for example, an antibody or a fragment thereof, anticalin, an N-recognition protein polypeptide (e.g., an ATP-dependent Clp protease adaptor polypeptide (ClpS)), an aptamer, etc., and variants or homologs thereof.

[0101] In some cases, the components of the conjugate-bound polypeptide include a polypeptide or a portion thereof obtained by fragmentation of a polypeptide from a variety of polypeptides. In some cases, the components of the conjugate-bound polypeptide include one or more amino acid residues. In some cases, the components of the conjugate-bound polypeptide include post-translational modifications. In some cases, the components of the conjugate-bound polypeptide include modified amino acid residues (e.g., modified NTAA or modified CTAA residues).

[0102] In some cases, one or more auxiliary sequences (e.g., 1, 2, 3, 4, 5 or more auxiliary sequences) may include one or more identifier sequences or their complements (e.g., 1, 2, 3, 4, 5 or more identifier sequences or their complements), one or more spacer sequences or their complements (e.g., 1, 2, 3, 4, 5 or more spacer sequences or their complements), one or more sequencing primer sequences or their complements (e.g., 1, 2, 3, 4, 5 or more sequencing primer sequences or their complements), or any combination thereof.

[0103] In some cases, one or more identifier sequences may include one or more sample barcode sequences (e.g., 1, 2, 3, 4, 5 or more sample barcode sequences or their complements), one or more bead barcode sequences (e.g., 1, 2, 3, 4, 5 or more bead barcode sequences or their complements), one or more unique molecular identifier (UMI) sequences (e.g., 1, 2, 3, 4, 5 or more UMI sequences or their complements), or any combination thereof.

[0104] In some cases, multiple nucleic acid sequences may include at least 5,000, 10,000, 20,000, 30,000, 40,000, 50,000, 60,000, 70,000, 80,000, 90,000, 100,000, 200,000, 300,000, 400,000, 500,000, 600,000, 700,000, 800,000, 900,000, 1M or more nucleic acid sequences.

[0105] In some cases, multiple peptides, polypeptides, polypeptides or fragments thereof may include at least 100, 500, 1,000, 1,500, 2,000, 2,500, 5,000, 7,500, 10,000 or more than 10,000 different polypeptides.

[0106] In some cases, the method may further include determining the nucleic acid sequence of multiple nucleic acid sequences. In some cases, the nucleic acid sequence can be determined by performing DNA sequencing. In some cases, DNA sequencing can be performed using, for example, a next-generation DNA sequencer.

[0107] exist Figure 3A At step 304A, one or more binder identifiers are assigned to each encoder barcode sequence in a series of encoder barcode sequences in each of the multiple nucleic acid sequences (and auxiliary identifiers are assigned to at least one of one or more auxiliary sequences) to generate multiple binder identifier strings.

[0108] In some cases, the assignment of the binder identifier to the encoder barcode sequence can be based on the known construction of nucleic acid sequences among multiple nucleic acid sequences (e.g., known patterns of sequencing primer sequences, beaded barcode sequences, binder barcode sequences, spacer sequences, etc.; see, for example...). Figure 6 and Figure 8 This includes a lookup table (LUT) for the barcode type. In some cases, the assignment of sub-identifiers to encoder barcode sequences may further include error correction based on an enumeration of error types to account for errors associated with the encoding process used to generate each encoder barcode sequence in a series of encoder barcode sequences. In some cases, considering errors associated with the process used to generate each encoder barcode sequence in a series of encoder barcode sequences may include determining: (i) a first probability of reading a specified barcode sequence of a specified underlying polypeptide component, and (ii) a second probability of successfully transferring from one polypeptide component to the next during the encoding process.

[0109] In some cases, the assignment of binder identifiers to encoder barcode sequences may include using a probabilistic model to predict the sequence of binder identifiers for each of a plurality of nucleic acid sequences. In some cases, for example, the probabilistic model may include a Hidden Markov Model (HMM). In some cases, the HMM may be trained using one or more training datasets (e.g., 1, 2, 3, 4, 5, or more than 5 training datasets) that include tag pairs of binder identifiers and nucleic acid sequences. In some cases, the HMM may be trained using an iterative expectation-maximization (EM) algorithm to determine a set of model parameters that maximize the probability of correctly predicting the sequence of binder identifiers for each of a plurality of nucleic acid sequences. In some cases, the expectation-maximization (EM) algorithm may include the Baum-Welch algorithm.

[0110] exist Figure 3AAt step 306A, multiple identifier strings are converted into multiple peptide reads (e.g., probabilistic peptide reads) based on, for example, the order of binding of binders in a coding assay (e.g., the order of binder identifiers in a given binder identifier string) and the determined binding characteristics of the binders (e.g., known or most probable components of the peptide or polypeptide fragment that a given binder may bind to (e.g., amino acid residues, modified amino acid residues, post-translational modifications, etc.)), which correspond to the binder identifier assigned to the given identifier string.

[0111] In some cases, at least two peptide reads (e.g., at least 2, 3, 4, 5, 10, 20, 30, 40, 50 or more than 50 nucleic acid sequences) are generated for at least one nucleic acid sequence among a plurality of nucleic acid sequences.

[0112] exist Figure 3A At step 308A, multiple peptide reads are provided as input to a computer model, which is configured to assign peptide reads to the amino acid sequence of a polypeptide and output the identity, partial identity, and / or quantity of at least one polypeptide present in the sample. In some cases, the computer model may output partial identity of at least one polypeptide, for example, where it is known that a conjugate binds to a polypeptide, a related polypeptide, or a family of homologous polypeptides.

[0113] In some cases, the computer model may be configured to identify a unique polypeptide signature among multiple peptide reads as part of assigning peptide reads to amino acid sequences. In some cases, a given unique polypeptide signature may include a set of peptide reads associated with a single polypeptide. In some cases, the computer model may be trained using a training dataset comprising peptide read data from one or more isolated polypeptide samples processed using the same sample preparation protocol used for processing multiple polypeptides. In some cases, the training dataset may further include peptide read data from background samples processed using the same sample preparation protocol used for processing multiple polypeptides. In some cases, background samples may include, for example, plasma samples, urine samples, saliva samples, or cell extract samples. In some cases, the samples may be cell extract samples, and cell extract samples may include, for example, mammalian cell extracts, plant cell extracts, fungal cell extracts, or bacterial cell extracts. In some cases, the computer model may be further configured to correct the quantity output of at least one polypeptide using a correction factor calculated from the training dataset.

[0114] In some cases, a computer model may be configured to map each of a plurality of peptide reads to a specific polypeptide as part of assigning peptide reads to amino acid sequences. In some cases, mapping peptide reads to a specific polypeptide may include: i) generating a set of k-mer fragments of the peptide reads; ii) determining the probability that a given k-mer fragment in the set belongs to a specific polypeptide based on a previously determined probability distribution; and iii) assigning the peptide read to the specific polypeptide based on the determined probability of the set of k-mer fragments. In some cases, the probabilistic model is trained on pre-determined (e.g., empirically determined) assay parameter data. In some cases, for each barcode sequence in a series of encoder barcode sequences, the empirically determined assay parameter data may include: the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of the corresponding polypeptide bound by the binder. In some cases, for each potential N-terminal amino acid residue in a polypeptide, the empirically determined assay parameter data may include: the probability of successfully cleaving an N-terminal amino acid residue during a cyclic process for encoding multiple polypeptides in multiple nucleic acid sequences. In some cases, mapping peptide reads from multiple peptide reads to a specific peptide may include providing multiple peptide reads as input to a computer model, wherein the computer model is configured to assign a given peptide read to two or more specific peptides in a fractional manner.

[0115] In some cases, the computer model can be trained on a training dataset comprising a set of simulated peptide reads generated based on empirically determined assay parameters for a given input distribution of a peptide. These assay parameters include: i) the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of the corresponding peptide bound by the binder, for each barcode sequence in a series of encoder barcode sequences; and ii) the probability of successfully cleaving an N-terminal amino acid residue for each potential N-terminal amino acid residue in the peptide during a cyclic process used to encode multiple peptides in multiple nucleic acid sequences. In some cases, the set of simulated peptide reads may include at least 100,000, 500,000, 1M, 5M, 10M, 25M, 50M, 75M, 100M, 200M, 300M, 400M, or 500M simulated peptide reads.

[0116] In some cases, the computer model can be further configured to output confidence intervals for partial identity of at least one peptide. In some cases, the stringency level of the output confidence interval can be selected by the user. In some cases, for example, the stringency level of the confidence interval corresponds to a confidence level of 90%, 95%, 98%, or 99%.

[0117] In some cases, the computer model may include, for example, a trained artificial neural network model, a trained deep learning model, a trained random forest model, or a trained support vector machine.

[0118] In some cases, the method may further include performing an iterative expectation-maximization (EM) process to refine the output of the quantity of at least one polypeptide among a plurality of polypeptides. In some cases, for example, the iterative EM process may repeatedly include: i) finding the optimal allocation of the binder to each barcode sequence in the nucleic acid sequence based on the current estimate of which polypeptides are present among the plurality of polypeptides; ii) finding an updated optimal estimate of which polypeptides are present among the plurality of polypeptides based on the optimal allocation of the binder to each different barcode sequence in the nucleic acid sequence; and iii) determining the quantity of at least one polypeptide present among the plurality of polypeptides based on the updated optimal estimate of which polypeptides are present among the plurality of polypeptides. In some cases, steps (i) to (iii) are repeated until the difference between the estimated quantity of at least one polypeptide determined in one iteration and the next is less than a specified threshold, or the sum of the squared differences between iterations is less than a specified threshold, or until a specified maximum number of iterations has been reached. In some cases, the specified threshold for the difference between the quantity of at least one polypeptide determined in one iteration and the next may be, for example, a difference of 5%, 1%, 0.1%, or 0.01%. In some cases, the maximum number of iterations can be, for example, 16, 32, 64, 128, or 256 iterations.

[0119] exist Figure 3A At step 310A, peptide read data and / or identification and quantification data may optionally be provided as input to any of a variety of application-specific analytical tools (e.g., software packages for statistical comparison of diseased samples against normal samples).

[0120] This article also describes methods (e.g., computer-implemented methods) for generating multiple coding nucleic acid sequences based on known amino acid sequences of multiple input peptides or peptide fragments. Such methods can be used, for example, to generate sequences for... Figure 3A or Figure 3B The training and / or validation data for the nucleic acid decoding process shown are useful.

[0121] In some cases, such methods may include: a) receiving (e.g., at one or more processors of a system configured to implement the method) amino acid sequences of multiple peptides; b) generating multiple peptide sequences based on the amino acid sequences; c) generating multiple peptide reads based on the multiple peptide sequences; d) converting each of the multiple peptide reads into an identifier string based on the order of multiple binders and determined binding characteristics, wherein each of the multiple binders binds to a component of the peptide and corresponds to a binder identifier, and wherein the binder identifier corresponds to a barcode sequence; e) providing the multiple identifier strings as input to a trained model, wherein the trained model is configured to convert the identifier strings into nucleic acid sequences and output multiple nucleic acid sequences corresponding to multiple peptides, the conversion being based on a set of probabilities of multiple binders binding to components of the peptides in the multiple peptides.

[0122] In some cases, the method may further include comparing multiple nucleic acid sequences output by a computer model with multiple nucleic acid sequences determined by subjecting multiple peptides to reverse translation assays (i.e., coding assays) and sequencing the resulting extended recording tags. In some cases, the computer model may include a statistical model. In some cases, the computer model may include a trained machine learning model.

[0123] Figure 3B Another non-limiting example of process 300B for processing barcoded nucleic acid sequences to identify peptides and / or polypeptides represented therein is provided. Process 300B may be performed, for example, using one or more electronic devices implementing a software platform. In some instances, process 300B is performed using a client-server system, and the blocks of process 300B are divided between server and client devices in any way. In other instances, the blocks of process 300B are divided between a server and multiple client devices. Therefore, although parts of process 300B are described herein as being performed by a specific device of a client-server system, it should be understood that process 300B is not limited thereto. In other instances, process 300B is performed using only client devices or only multiple client devices. In process 300B, some blocks are optionally combined, the order of some blocks is optionally changed, and some blocks are optionally omitted. In some instances, additional steps may be combined with process 300B. Therefore, the operations shown (and described in more detail below) are exemplary in nature and should therefore not be considered limiting.

[0124] exist Figure 3BAt step 302B, raw NGS data of multiple peptides, polypeptides, polypeptides or fragments thereof (e.g., peptides, polypeptides or polypeptides extracted from a sample) encoded in multiple nucleic acid sequences (i.e., extended record tag sequences) are received (e.g., by one or more processors of a system configured to implement the disclosed method).

[0125] As mentioned above Figure 3A The raw NGS data is processed to create and output a sequence read file (e.g., a FASTQ file) that includes nucleic acid sequence reads corresponding to a plurality of nucleic acid sequences (corresponding to extended record tag sequences), wherein each of the plurality of nucleic acid sequences contains a series of encoder barcode sequences and optionally one or more auxiliary sequences, and wherein each of the series of encoder barcode sequences corresponds to a binder of a component of a given polypeptide that binds to a plurality of polypeptides (described elsewhere herein).

[0126] exist Figure 3B At step 304B, a barcode-type lookup table (LUT) is used to assign one or more binder identifiers to each encoder barcode sequence in a series of encoder barcode sequences in each of the multiple nucleic acid sequences (and to assign auxiliary identifiers to at least one of one or more auxiliary sequences) to generate multiple identifier strings.

[0127] Lookup tables (LUTs) are used to quickly assign binder identifiers to different encoder barcode sequences identified in sequence reads based on known constructs of nucleic acid sequences (extended record tag sequences), as referenced above. Figure 3A The assignment of sub-identifiers to encoder barcode sequences may, in some cases, further include error correction based on an enumeration of error types to account for errors associated with the encoding process used to generate each encoder barcode sequence in a series of encoder barcode sequences. In some cases, considering errors associated with the process used to generate each encoder barcode sequence in a series of encoder barcode sequences may include determining: (i) a first probability of reading a specified barcode sequence of a specified underlying polypeptide component, and (ii) a second probability of successfully transferring from one polypeptide component to the next during the encoding process.

[0128] Alternatively, in Figure 3BAt step 306B, the assignment of the binder identifier to the encoder barcode sequence may include using a probabilistic model to predict the sequence of the binder identifier for each of a plurality of nucleic acid sequences. In some cases, for example, the probabilistic model may include a Hidden Markov Model (HMM). In some cases, the HMM may be trained using one or more training datasets (e.g., 1, 2, 3, 4, 5, or more than 5 training datasets) that include tag pairs of binder identifiers and nucleic acid sequences. In some cases, the HMM may be trained using an iterative expectation-maximization (EM) algorithm to determine a set of model parameters that maximize the probability of correctly predicting the sequence of the binder identifier for each of a plurality of nucleic acid sequences. In some cases, the expectation-maximization (EM) algorithm may include the Baum-Welch algorithm.

[0129] In some cases, the Hidden Markov Model (HMM) / Baum-Welch method may be computationally slower than the LUT method, but it may be more accurate in correctly assigning barcode (and auxiliary sequence) identifiers.

[0130] exist Figure 3B At step 308B, output the multiple identifier strings generated in step 304B or step 306B.

[0131] exist Figure 3B At step 310B, multiple identifier strings are optionally converted into multiple peptide reads based on, for example, the order of the binders (e.g., the order of the binder identifiers in the given identifier string) and a determined binding characteristic (e.g., a known component of the peptide or polypeptide fragment bound by the given binder (e.g., amino acid residues, modified amino acid residues, post-translational modifications, etc.)), the binder corresponding to the binder identifier assigned to the given identifier string.

[0132] In some cases, at least two peptide reads (e.g., at least 2, 3, 4, 5, 10, 20, 30, 40, 50 or more than 50 nucleic acid sequences) are generated for at least one nucleic acid sequence among a plurality of nucleic acid sequences.

[0133] exist Figure 3BAt step 312B, multiple binder identifier strings or peptide reads (if converted from binder identifier strings) are provided as input to the computer model, and the batch analysis of the binder identifier string or peptide read data is mapped to peptide identification and quantification of at least one peptide. That is, the computer model can be configured to identify a unique peptide signature profile among multiple binder identifier strings or peptide reads as part of assigning the binder identifier strings or peptide reads to amino acid sequences. In some cases, a given unique peptide signature profile may include a set of binder identifier strings or peptide reads associated with a single peptide. In some cases, the computer model may be trained using a training dataset comprising binder identifier string or peptide read data of one or more isolated peptide samples processed using the same sample preparation protocol used for processing multiple peptides. In some cases, the training dataset may further include binder identifier string or peptide read data of background samples processed using the same sample preparation protocol used for processing multiple peptides. In some cases, background samples may include, for example, plasma samples, urine samples, saliva samples, or cell extract samples. In some cases, the sample may be a cell extract sample, and the cell extract sample may include, for example, mammalian cell extracts, plant cell extracts, fungal cell extracts, or bacterial cell extracts. In some cases, the computer model may be further configured to correct the quantity output of at least one peptide using a correction factor calculated from the training dataset.

[0134] Alternatively, in Figure 3BAt step 314B, multiple binder identifier strings or peptide reads are provided as input to a computer model, and peptide identification and quantification are performed based on read-level analysis of the binder identifier string or peptide read data, mapping to at least one peptide. That is, the computer model can be configured to map each of the multiple binder identifier strings or peptide reads to a specific peptide as part of assigning the binder identifier string or peptide read to an amino acid sequence. In some cases, mapping the multiple binder identifier strings or peptide reads to a specific peptide may include: i) generating a set of k-mer fragments of the binder identifier strings or peptide reads; ii) determining the probability that a given k-mer fragment in the set belongs to a specific peptide based on a previously determined probability distribution; and iii) assigning the binder identifier string or peptide read to a specific peptide based on the determined probability of the set of k-mer fragments. In some cases, the probabilistic model is trained on assay parameter data determined empirically. In some cases, for each barcode sequence in a series of encoder barcode sequences, empirically determined assay parameters may include: the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of the corresponding polypeptide bound by the conjugate. In some cases, for each potential N-terminal amino acid residue in a polypeptide, empirically determined assay parameters may include: the probability of successfully cleaving the N-terminal amino acid residue during a cyclic process used to encode multiple polypeptides in multiple nucleic acid sequences. In some cases, mapping multiple conjugate identifier strings or peptide reads to specific polypeptides may include providing multiple conjugate identifier strings or peptide reads as input to a computer model, wherein the computer model is configured to fractionally assign a given conjugate identifier string or peptide read to two or more specific polypeptides.

[0135] The output of the computer model in batch analysis and / or read-level analysis methods may include the identification and / or quantification of at least one peptide, polypeptide, or peptide present in the original sample. In some cases, the quantification of the peptide, polypeptide, or peptide present in the sample may be specified, for example, in parts per million (ppm), such as the number of peptide, polypeptide, or peptide molecules of a given species present in every million peptide, polypeptide, or peptide molecules detected. In some cases, the output of the computer model may include determining, for example, that a binder identifier string or peptide read X originates from polypeptide Y, and that polypeptide Y is present at Z parts per million.

[0136] In some cases, the computer model can be trained on a training dataset comprising a set of simulated binder identifier strings generated based on empirically determined assay parameter data for a given input distribution of a peptide. The assay parameter data includes: i) the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of the corresponding peptide bound by the binder, for each barcode sequence in a series of encoder barcode sequences; and ii) the probability of successfully cleaving an N-terminal amino acid residue for each potential N-terminal amino acid residue in the peptide during a cyclic process used to encode multiple peptides in multiple nucleic acid sequences. In some cases, the set of simulated binder identifier strings may include at least 100,000, 500,000, 1M, 5M, 10M, 25M, 50M, 75M, 100M, 200M, 300M, 400M, or 500M simulated binder identifier strings.

[0137] In some cases, the computer model can be further configured to output confidence intervals for partial identity of at least one peptide. In some cases, the stringency level of the output confidence interval can be selected by the user. In some cases, for example, the stringency level of the confidence interval corresponds to a confidence level of 90%, 95%, 98%, or 99%.

[0138] In some cases, the computer model may include, for example, a trained artificial neural network model, a trained deep learning model, a trained random forest model, or a trained support vector machine.

[0139] In some cases, the method may further include performing an iterative expectation-maximization (EM) process to refine the output of the quantity of at least one polypeptide among a plurality of polypeptides. In some cases, for example, the iterative EM process may repeatedly include: i) finding the optimal allocation of the binder to each barcode sequence in the nucleic acid sequence based on a current estimate of which polypeptides are present among the plurality of polypeptides; ii) finding an updated optimal estimate of which polypeptides are present among the plurality of polypeptides based on the optimal allocation of the binder to each different barcode sequence in the nucleic acid sequence; and iii) determining the quantity of at least one polypeptide present among the plurality of polypeptides based on the updated optimal estimate of which polypeptides are present among the plurality of polypeptides. In some cases, steps (i) through (iii) are repeated until the difference between the quantity of at least one polypeptide determined in one iteration and the next is less than a specified threshold, or until a specified maximum number of iterations has been reached. In some cases, the specified threshold for the difference between the quantity of at least one polypeptide determined in one iteration and the next may be, for example, a difference of 10%, 8%, 6%, 4%, or 2%. In some cases, the maximum number of iterations can be, for example, 5, 10, 15, 20, 25, 30, 35, 40, 45, or 50 iterations.

[0140] exist Figure 3B At step 316B, peptide read data and / or identification and quantification data may optionally be provided as input to any of a variety of application-specific analytical tools (e.g., software packages for statistical comparison of diseased samples against normal samples).

[0141] As mentioned above Figure 3A In some cases, computer models can be trained and / or validated using a combination of experimental and / or simulated data (e.g., multiple coding nucleic acid sequences generated in a computer based on known amino acid sequences of multiple input peptides or peptide fragments).

[0142] Figure 4A Another non-limiting example of a process 400A for processing barcoded nucleic acid sequences to identify peptides represented therein is provided. Process 400A may be performed in part as described above for processes 300A and 300B. In process 400A, step 402A is identical to step 302B in process 300B; and step 404A is identical to step 304B or step 306B in process 300B (see description of steps 302B, 304B, and 306B above).

[0143] exist Figure 4A At step 406A, a set of simulated probability-binding sub-identifier strings is generated in a computer for each polypeptide and / or polypeptide fragment based on predetermined parameters of the coding assay (see also Example 8). When the number of polypeptides to be analyzed is limited (e.g., less than 100), the probability distribution of the polypeptides to be analyzed can be generated for polypeptides used in the coding assay (e.g., trypsin digest of proteins from one or more biological samples). When the number of polypeptides to be analyzed exceeds a certain threshold (e.g., greater than 100, greater than 1000, or greater than 10000), the computational power required to generate the probability distribution of the polypeptides to be analyzed can be impaired. Therefore, in order to analyze a large number of polypeptides simultaneously (e.g., analyzing the entire eukaryotic proteome), each polypeptide can be broken down into k-mer fragments as described in Example 8, which reduces the computational power requirements.

[0144] In a preferred embodiment, predetermined parameters for the coding assay are obtained by performing multiple coding assays using a set of known peptide analytes. For example, when each of the binding molecules in the set used in the coding assay is specific to a particular P1-P2 residue of the peptide analyte, a set of model peptides having all possible combinations of P1-P2 can be used for the coding assay (see, for example, Figures 16A to 16G(Heatmap encoded data). In some embodiments, predetermined parameters of the encoding assay are obtained experimentally and further used to train a computer model that uses them as input parameters. In some embodiments, the parameters of the encoding assay may not be predetermined but are computed within a machine learning (ML) model, which is performed when encoded data from a large number of model peptides (such as genetically encoded peptides representing multiple different amino acid sequences) are input into the ML model (see Example 10 for more details), and the model output includes data from multiple nucleic acid sequences. Millions of encoding assays can be performed in parallel to provide a sufficient dataset for training the ML model.

[0145] In some embodiments, in addition to training the ML model to recognize the components of a peptide analyte via binders, training can also provide information on the efficiency of peptide component cleavage in the coding assay, such as the efficiency of cleavage of N-terminal amino acid residues after each coding cycle, which may depend on adjacent amino acid residues (see Example 7). Training can be performed by feeding the ML model experimental output (i.e., data including multiple nucleic acid sequences) generated from multiple coding assays using a predetermined peptide analyte (such as peptides containing all possible NTAA residues at a point during a cleavage event) as input, and analyzing the encoded binding events to detect the presence or absence of cleavage events dependent on adjacent amino acid residues. For example, depending on the cleavage method (e.g., chemical cleavage or enzymatic cleavage), the efficiency of NTAA cleavage can vary depending on the specific NTAA cleaved. In various embodiments, the cleavage efficiency can vary from 70% to about 100%.

[0146] In some embodiments, predetermined parameters of the coding assay include: i) the probability of correctly or incorrectly assigning one or more binder identifiers to at least one component of a given polypeptide or polypeptide fragment based on the determined binding characteristics of the binders used in the coding assay; and ii) optionally, the probability of successfully cleaving at least one component of a given polypeptide or polypeptide fragment after a binder binds to at least one component in the coding assay. In a preferred embodiment, the predetermined parameters include both elements i) and ii).

[0147] exist Figure 4AAt step 408A, using the simulation assay method disclosed in step 406A and the probability calculated for each simulated binder identifier string, the probability of each of the multiple binder identifier strings generated in step 404A is calculated (i.e., the probability that a given binder identifier string is generated from one or more polypeptides or fragments thereof). This enables the experimental coding data (nucleic acid sequence) to be correlated or directly matched with a specific polypeptide or fragment thereof present in the analyzed sample (or multiple samples) with a certain probability. The process will be able to determine at least partial identity and / or quantity of at least one polypeptide among multiple polypeptide analytes.

[0148] In a preferred embodiment, as Figure 4A The result of step 408A, the probability of each simulated binder identifier string generated from a given polypeptide or polypeptide fragment, is calculated using predetermined assay parameters (see Example 8 for an exemplary calculation). The data generated during the simulation can be reorganized such that, for each simulated binder identifier string, the probability that the identifier string originates from a specific polypeptide is determined.

[0149] In process 400A, step 410A includes calculating a probability score based on one or more simulated binder identifier strings, and matching each binder identifier string with one or more simulated binder identifier strings. Step 410A includes outputting data relating to at least partial identity and / or quantity of at least one of a plurality of polypeptides, similar to step 316B in process 300B (see description above).

[0150] Figure 4B Another non-limiting example of a process 400B for processing barcoded nucleic acid sequences to identify polypeptides represented therein is provided. Process 400B may be performed in part as described above for processes 300A and 300B. In process 400B, step 402B is identical to step 302B in process 300B; and step 404B is identical to step 304B or step 306B in process 300B (see description of steps 302B, 304B, and 306B above).

[0151] exist Figure 4B At step 406B, each binder identifier string is converted into multiple peptide reads based on the determined binding characteristics of the binding portion, the binding portion corresponding to the binder identifier assigned in the given binder identifier string; and a probability score for each peptide read is generated based on the probability that the given peptide read is generated from the given binder identifier string (see Example 9 for further details).

[0152] exist Figure 4BAt step 408B, peptide reads are filtered out from each of the multiple peptide reads generated for a given binder identifier string, based on (i) the probability score of each peptide read and (ii) the likelihood of generation from multiple peptides. This step is optional and can be performed iteratively once or multiple times, while inferring the amino acid sequence of the peptide from the binder identifier in the binder identifier string. This step simplifies the association between binder identifier strings generated from coding assay data and the amino acid sequences of peptides from multiple peptides by filtering out low-probability binder identifier strings that are unlikely to be generated from the amino acid sequences of multiple peptides that may be present in one or more samples analyzed. In some embodiments, the filtering step may be based on a predefined threshold (see Example 9 for further details).

[0153] Finally, step 410B includes outputting data relating to at least partial identity and / or quantity of at least one of a plurality of polypeptides, similar to step 316B in process 300B (see description above).

[0154] Machine learning methods

[0155] Any of a variety of machine learning methods and algorithms (where machine learning models, as referred to herein, include trained machine learning algorithms) can be used to implement the disclosed methods. For example, the machine learning models / algorithms employed may include supervised learning models / algorithms, unsupervised learning models / algorithms, semi-supervised learning models / algorithms, deep learning models / algorithms, or any combination thereof, as discussed in more detail below. In some cases, one or more machine learning models (e.g., 1, 2, 3, 4, 5, 6, or more than 6 machine learning models) may be used to implement the disclosed methods.

[0156] Examples of machine learning algorithms that may be used include, but are not limited to, artificial neural networks, deep neural networks, deep recurrent neural networks, deep convolutional neural networks, Gaussian process regression, logistic model tree, random forest, fuzzy classifier, decision tree, hierarchical clustering, k-means clustering, fuzzy clustering, deep Boltzmann machine learning, or any combination thereof.

[0157] Supervised learning models / algorithms: Supervised learning models include trained algorithms that rely on using labeled training data to infer a relationship between a set of one or more input features (e.g., multiple binder identifier strings) and an output prediction (e.g., identification and / or quantification of at least one peptide represented by multiple binder identifier strings). The training data includes one or more sets of paired training instances (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or more than 10 sets), where each instance includes a set of input features (e.g., a set of binder identifier strings) and a corresponding output (e.g., identification and / or quantification of a peptide corresponding to a set of binder identifier strings).

[0158] Unsupervised learning models / algorithms: Unsupervised learning algorithms are algorithms used to make inferences from training datasets consisting of input datasets (e.g., a set of binder identifier strings) that are not paired with labeled output data (e.g., identification and / or quantification of peptides corresponding to a set of binder identifier strings). A common example of an unsupervised learning algorithm is clustering analysis, which is often used in exploratory data analysis to find hidden patterns or groupings in multidimensional datasets. Other examples of unsupervised learning algorithms include, but are not limited to, artificial neural networks, association rule learning algorithms, hierarchical clustering algorithms, matrix factorization methods, dimensionality reduction methods, or any combination thereof.

[0159] Semi-supervised learning models / algorithms: Semi-supervised learning algorithms are algorithms that use both labeled and unlabeled training data for training (typically using a relatively small amount of labeled data and a larger amount of unlabeled data).

[0160] Artificial Neural Networks and Deep Learning Models / Algorithms: Artificial neural networks (ANNs) and deep learning algorithms are algorithms inspired by the structure and function of the human brain. Specifically, deep learning algorithms are large artificial neural networks comprising many layers coupled with "nodes," which can be used to map input feature data to, for example, classification decisions. In some cases, the machine learning models / algorithms used to implement the disclosed methods and systems may be artificial neural networks (ANNs) or deep learning models / algorithms, including any type of neural network model known to those skilled in the art, such as feedforward neural networks, radial basis function networks, recurrent neural networks, or convolutional neural networks, etc. In some cases, the disclosed methods and systems may employ pre-trained ANN or deep learning models. In some embodiments, the disclosed methods and systems may employ continuously learned ANN or deep learning models, wherein the model is periodically or continuously updated based on new training data provided, for example, by a single local operating system, multiple local operating systems, or multiple geographically distributed operating systems.

[0161] Artificial neural networks (ANNs) generally consist of interconnected groups of nodes organized into multiple layers. For example, an ANN architecture may include at least an input layer, one or more hidden layers, and an output layer. An ANN can include any total number of layers and any number of hidden layers, where the hidden layers act as trainable feature extractors capable of mapping a set of input data to preferred output values ​​or a set of output values. Each layer of a neural network includes multiple nodes (or neurons). Nodes receive input data (e.g., combined with sub-identifier string data) directly from the input data node or from the output of nodes in previous layers, and perform specific operations, such as summation. In some cases, the connections from the input to the node are associated with weights (or weighting factors). In some cases, a node may, for example, be assigned weights to the input X. i and its associated weight W i The sum of the products of all pairs is calculated. In some cases, the weighted sum is offset by a bias b. In some cases, the neuron's output can be gated using a threshold or an activation function f, which can be a linear or nonlinear function. The activation function can be, for example, the Modified Linear Unit (ReLU) activation function or other functions such as the saturated hyperbolic tangent function, identity function, binary step function, logistic function, arcTan function, softsign function, parametric modified linear unit function, exponential linear unit function, softPlus function, curved identity function, softExponential function, Sinusoid function, Sine function, Gaussian function, sigmoid function, or any combination thereof.

[0162] Weighting factors, bias values, and thresholds, or other computational parameters of the neural network, can be "taught" or "learned" during the training phase using one or more sets of training data. For example, parameters can be trained using input data from the training dataset and gradient descent or backpropagation methods such that the output values ​​computed by the ANN (e.g., identification and / or quantification of at least one peptide) are consistent with the instances included in the training dataset. The adjustable parameters of the model can be obtained from the backpropagation neural network training process, which may or may not use the same hardware used to perform the disclosed method.

[0163] Generally, the number of nodes in the input layer of an ANN or deep learning model / algorithm (which can take in, for example, large amounts of data or data combined with sub-identifier strings) can range from about 10 to about 10,000 nodes. In some cases, the number of nodes used in the input layer can be at least 10, at least 50, at least 100, at least 200, at least 300, at least 400, at least 500, at least 600, at least 700, at least 800, at least 900, at least 1000, at least 2000, at least 3000, at least 4000, at least 5000, at least 6000, at least 7000, at least 8000, at least 9000, or at least 10,000 nodes, or any number of nodes within that range.

[0164] In some cases, the total number of layers (including input and output layers) used in an ANN or deep learning model / algorithm can range from about 3 to about 20 layers or more. In some cases, for example, the total number of layers is at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 15 or at least 20 layers, or any number of layers within that range.

[0165] In some cases, the total number of learnable or trainable parameters (e.g., weighting factors, biases, or thresholds) used in an ANN or deep learning model / algorithm can range from about 1 to about 1,000,000,000. In some cases, the total number of learnable parameters can be at least 1, at least 10, at least 100, at least 500, at least 1,000, at least 2,000, at least 3,000, at least 4,000, at least 5,000, at least 6,000, at least 7,000, at least 8,000, at least 9,000, at least 10,000, at least 100,000, at least 1,000,000, at least 10,000,000, at least 100,000,000, or at least 1,000,000,000 parameters, or any number of parameters within that range.

[0166] In some embodiments, an ANN may consist of many more nodes / layers / trainable parameters, such as an ANN having millions or billions of parameters in a properly designed model.

[0167] Hidden Markov Models (HMMs): HMMs are probabilistic models that describe the evolution of observable events (symbols) over time, which depend on unobservable internal factors (hidden states) (see, for example, Yoon (2009), “Hidden Markov Models and their Applications in Biological Sequence Analysis”, Current Genomics 10:402-415; Tamposis et al. (2019), “Semi-Supervised Learning of Hidden Markov Models for Biological Sequence Analysis”, Bioinformatics 35(13);2208-2215). They have been widely used in applications such as pattern recognition, speech recognition, digital communication, and computational analysis of biological sequences (e.g., gene prediction, sequence alignment, base interpretation, modeling DNA sequencing errors, and predicting peptide secondary structures).

[0168] Suppose the system being modeled is governed by a Markov process, which is a stochastic process consisting of a sequence of possible events, where the probability of each event depends only on the underlying hidden states obtained from previous events. For the sequence of events, in the set O of observations x = x1, x2, x3, …x L Furthermore, the underlying sequence of states y = y1, y2, y3, …y is hidden in the set of states S. L Assume the sequence of hidden states evolves in a time-homogeneous manner, meaning the probability of entering state j at the next time point depends only on the current state I and does not change over time. This fixed probability of transitioning from state I to state j...

[0169]

[0170] (For all states I, j ∈ S and for all n ≥ 1) is called the transition probability. For the initial state y1, the initial state probability is expressed as... (For all I ∈ S).

[0171] The nth observation of the event will be x = x n The probability depends only on the underlying state y = y n = I,

[0172]

[0173] (For all possible observations x ∈ O, all states I ∈ S, and all n ≥ 1), and is called the emission probability of x at state i. In summary, the set of three probability measures t(I, j), π(i), and e(x|i) (denoted as Θ) completely specifies the Hidden Markov Model.

[0174] Hidden Markov Models can be trained using supervised, semi-supervised, or unsupervised learning methods. For example, in supervised machine learning methods, training data including instances of input data (e.g., a set of binder identifier strings) and accompanying labels corresponding to different output categories (e.g., different peptides) are provided as input to the model, and a set of model parameters is determined that maximizes the joint probability of correctly mapping the input data to the appropriate output category.

[0175] Hidden Markov Models (HMMs) are typically trained using an iterative expectation-maximization (EM) algorithm. Each iteration consists of an "estimation" step and a "maximization" step. In the "maximization" step, each set of input data values ​​(including one set of input data values ​​for the observation vector X) is compared with a state S in the model to maximize the likelihood measure. In the "estimation" step, for each state S, the following are estimated: (i) the parameters of the statistical model that the observation vector X is compared with state S, and (ii) the state transition probabilities. In subsequent iterations, the maximization step is repeated with the updated statistical model parameters. This process is repeated for a specified maximum number of iterations, or until the change in the likelihood measure from one iteration to the next is less than a specified threshold (i.e., the model converges to a stable solution). HMMs typically have a "starting" state marked by the first set of observations compared and a unidirectional topology such that once the system has transitioned from a given state, it does not return to that state.

[0176] Training Datasets: As mentioned above, the type of training data used to train machine learning models / algorithms for publicly available methods and systems will depend on, for example, whether a supervised or unsupervised approach is employed. In some cases, one or more training datasets may be used to train computer models in training phases different from the deployment phase. In some cases, training data may be continuously updated and used to update machine learning algorithms in real time. In some cases, training data may be stored in a training database residing on a local computer or server. In some cases, training data may be stored in a training database residing online or in the cloud.

[0177] Machine learning software: Any of the various commercial or open-source software packages, software languages, or software platforms known to those skilled in the art that can be used to implement publicly disclosed methods and systems of machine learning algorithms. Examples include, but are not limited to, Shogun (…). www.shogun -toolbox.org), Mlpack ( www.mlpack.rog), R (r-project.org), Weka ( www.cs .waikato.ac.nz / ml / weka / ), Python ( www.python .org) and / or MATLAB (MathWorks, Natick, MA, www.mathworks .com).

[0178] Systems and software for deriving the identity of encoded peptides

[0179] This paper also discloses a system designed to implement any of the publicly available bioinformatics methods for deriving the identity of encoded peptides. The system may include, for example, one or more processors, and a memory unit communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to: receive a plurality of nucleic acid sequences, wherein each of the plurality of nucleic acid sequences comprises a series of encoder barcode sequences and one or more auxiliary sequences, and wherein each of the series of encoder barcode sequences corresponds to a binder of a component of a polypeptide bound to a plurality of polypeptides; assign a binder identifier to each of the series of encoder barcode sequences and assign an auxiliary identifier to at least one of the one or more auxiliary sequences in each of the plurality of nucleic acid sequences to generate a plurality of identifier strings for the plurality of nucleic acid sequences; convert the plurality of identifier strings into a plurality of binder identifier strings based on the order of the binders and determined binding characteristics, wherein the binders correspond to binder identifiers assigned to a given identifier string; and provide the plurality of binder identifier strings as input to a computer model, wherein the computer model is configured to assign the binder identifier strings to amino acid sequences of polypeptides in a plurality of polypeptides and output partial identity and / or quantity of at least one polypeptide in a plurality of polypeptides.

[0180] In some cases, the disclosed system may further include sequencers, such as next-generation sequencers. Examples of next-generation sequencing platforms include, but are not limited to, Roche / 454's Genome Sequencing (GS) FLX system, Illumina / Solexa's Genome Analyzer (GA), Illumina's HiSeq® 2500, HiSeq® 3000, HiSeq® 4000 and NovaSeq® 6000 sequencing systems, Life / APG's Supported Oligonucleotide Ligation Detection (SOLiD) system, Polonator's G.007 system, Helicos BioSciences' HeliScope gene sequencing system, ThermoFisher Scientific's IonTorrent Genexus system, or Pacific Biosciences' PacBio® RS system.

[0181] Software including encoded instructions necessary for implementing any of the methods described herein is also included in this disclosure. For example, this document discloses a non-transitory computer-readable storage medium storing one or more programs comprising instructions that, when executed by one or more processors of a system, cause the system to: receive a plurality of nucleic acid sequences generated from an encoding assay, wherein each of the plurality of nucleic acid sequences comprises a series of encoder barcode sequences, and wherein each of the series of encoder barcode sequences corresponds to a conjugate that binds to a component of a polypeptide among a plurality of polypeptides in the encoding assay; assign conjugate identifiers to each of the series of encoder barcode sequences among the plurality of nucleic acid sequences to generate a plurality of identifier strings for the plurality of nucleic acid sequences; and provide the plurality of conjugate identifier strings as input to a computer model, wherein the computer model is configured to infer the amino acid sequence of a polypeptide among the plurality of polypeptides from the conjugate identifier strings and output data relating to the partial identity and / or quantity of at least one polypeptide among the plurality of polypeptides, the inference being based on a determined binding characteristic of a conjugate that corresponds to a conjugate identifier in a given identifier string.

[0182] The following paragraphs provide a non-restrictive list of known algorithms, each of which can be used to (i) calculate a probability score of the association between one or more binder identifiers in a binder identifier string and one or more amino acid sequences of a polypeptide; and / or (ii) infer the amino acid sequence of a polypeptide from a variety of polypeptides from the binder identifiers in the binder identifier string.

[0183] In some embodiments, a Bayesian network (BN) can be implemented to infer a polypeptide sequence from a given binder identifier string, wherein each binder identifier serves as a node in the network, nodes can be sequentially connected based on the order in which the binder identifier affects amino acid outcomes, and wherein a conditional probability table is defined, representing the likelihood of each possible amino acid identity outcome given one or more parent binder identifiers. Inference can optionally be performed on the BN using a variable elimination algorithm, a belief propagation algorithm, or a Gibbs sampling algorithm.

[0184] In another non-limiting embodiment, a Conditional Random Field (CRF) can be used to infer a polypeptide sequence from a given binder identifier string, modeling the conditional dependencies between output markers (i.e., amino acids in the polypeptide sequence) given input features (i.e., binder identifier transitions), and can optionally be trained using gradient descent with a finite-memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) optimization algorithm, stochastic gradient descent with an L2 regularization term optimization algorithm, average perceptron optimization algorithm, passive attack (PA) optimization algorithm, and / or adaptive weight vector regularization (AROW) optimization algorithm.

[0185] In another non-limiting embodiment, a Markov random field (MRF) can be implemented to predict polypeptide sequences from a string of binder identifiers. This may optionally involve: treating each position in the binder identifier string as a node; defining a potential for each node that represents the probability of producing a particular amino acid identity, optionally adding pairwise potentials to represent dependencies between consecutive binder identifiers (e.g., optionally including these joint probabilities in the model if adjacent binder identifiers have a higher probability of producing a pair of amino acid identities); and using inference techniques (including, but not limited to, confidence propagation or Gibbs sampling) to compute the most probable polypeptide sequence given the modeled potentials.

[0186] In another non-limiting embodiment, a Kalman filter (or linear quadratic estimation) algorithm may be implemented to infer a polypeptide sequence from a string of binder identifiers, wherein the Kalman filter is optionally tuned to incorporate the transition probabilities between one or more binder identifiers in the binder identifier string and the likelihood that they produce one or more amino acids in the polypeptide sequence. In some cases, the Kalman filter algorithm may involve defining a state model, where each state represents a hidden state of the polypeptide sequence linked to a binder identifier or a transition between binder identifiers, and may involve defining an observation model that correlates each binder identifier with the probability of amino acid identity in the polypeptide sequence. The Kalman filter components may involve: a state vector, where each element of the vector represents the probability of an amino acid in the polypeptide sequence at a certain position in the polypeptide sequence; a transition matrix, which models the transitions between binder identifiers in the binder identifier string, and wherein some transitions are optionally more likely than others based on known polypeptide sequences in the dataset; an observation matrix, which captures the probability of each binder identifier emitting a specific amino acid identity; and / or a covariance matrix, which defines the covariance of process noise (i.e., transition variability) and observation noise (i.e., uncertainty of emission probability). After processing all the binder identifiers in the binder identifier string, the Kalman filter can have an adjusted state vector representing the most likely polypeptide sequence.

[0187] In another non-limiting embodiment, implementing a particle filter (or sequence Monte Carlo) algorithm to infer a polypeptide sequence from a binder identifier string may involve: initializing particles, where each particle represents a possible polypeptide sequence outcome; one or more propagation or prediction steps, wherein new possible amino acid identities are generated by extending each particle based on the probability observed in the dataset of binder identifiers that generate amino acid identities; one or more weight update or likelihood calculation steps, wherein weights are assigned to each particle based on how well the new amino acid matches the probability of the aforementioned observation; resampling the particles based on their weights, optionally copying particles with high weights and discarding particles with low weights; optionally iterating through the propagation, weight update, and resampling steps for each binder identifier; and extracting polypeptide sequence predictions, wherein the particle with the highest weight may be considered the most likely polypeptide sequence, or the particles weighted in the top N may be considered to predict a range of most likely polypeptide sequences, where N is a positive integer greater than zero.

[0188] In another non-limiting embodiment, a Gaussian process (GP) algorithm can be implemented to infer polypeptide sequences from binder identifier strings, which may involve: representing each binder identifier string as a point in the input space, each binder identifier as a categorical or numerical feature, and representing one or more individual amino acids in the polypeptide sequence as points in the output space as categorical or numerical features; capturing the similarity between binder identifiers using a kernel function (e.g., a radial basis function or a Matern kernel function); training the GP algorithm with the probability of observations mapping one or more binder identifiers to one or more amino acids as a label for the GP; and running inference on the trained model, wherein, given a new binder identifier string, the identity of one or more amino acids is predicted by drawing from the GP posterior and optionally obtaining the average prediction and uncertainty (or confidence interval) of the prediction.

[0189] In another non-limiting embodiment, an autoregressive integral moving average (ARIMA) model is implemented to predict the next amino acid identity in a polypeptide sequence by taking into account previously determined amino acid identities in the polypeptide sequence and the binder identifier string.

[0190] In another non-limiting embodiment, a dynamic time warping (DTW) algorithm can be implemented to find the optimal alignment between the binder identifier string and the peptide sequence, taking into account inefficiencies in the coding assay steps, such as one or more skipped N-terminal modifications and / or one or more skipped NTAA modification cleavage steps and / or one or more unmodified NTAA cleavage steps.

[0191] In another non-limiting embodiment, a neural autoregressive model can be implemented to infer a polypeptide sequence from a binder identifier string, which involves autoregressively generating an amino acid sample in an output sequence given an input sequence that may involve a binder identifier string.

[0192] In another non-limiting embodiment, a variational autoencoder (VAE) algorithm can be implemented to infer polypeptide sequences from binder identifier strings. This algorithm can be trained to learn a latent spatial representation of a dataset that maps binder identifier strings to polypeptide sequences, and can be used to infer to generate polypeptide sequences given previously unseen binder identifier strings.

[0193] In another non-limiting embodiment, a generative adversarial network (GAN) algorithm can be implemented to infer polypeptide sequences from a binder identifier string. This algorithm uses a generator model (which takes the input binder identifier string and outputs a polypeptide sequence) and a discriminator model (which classifies the polypeptide sequence generated by the generator model as a true or false polypeptide sequence) within an adversarial framework to learn a probability distribution of amino acid identity given a binder identifier. The generator model can be trained to generate polypeptide sequences conditionally on the binder identifier string, minimizing the probability that the discriminator model correctly identifies its sequence as false, and the discriminator model is trained to maximize the probability of classifying a true polypeptide sequence as true and minimize the probability of classifying a false polypeptide sequence as true. Optionally, a conditional GAN ​​(cGAN) approach can be employed, where both the generator and discriminator are conditional on the binder identifier input (Mirza, M. (2014). Conditional generative adversarial nets. arXiv preprint arXiv:1411.1784.).

[0194] Computer devices and systems

[0195] Figure 5 Examples of computing devices according to one or more embodiments of the present disclosure are shown. Device 500 may be a host computer connected to a network. Device 500 may be a client computer or a server. Figure 5 As shown, device 500 can be any suitable type of microprocessor-based device, such as a personal computer, workstation, server, or handheld computing device (portable electronic device), such as a telephone or tablet. The device may include, for example, one or more of a processor 510, an input device 520, an output device 530, a memory 540, and a communication device 560. Input device 520 and output device 530 typically correspond to those described above, and they can be connected to or integrated with a computer.

[0196] Input device 520 may be any suitable device that provides input, such as a touch screen, keyboard or keypad, mouse or voice recognition device. Output device 530 may be any suitable device that provides output, such as a touch screen, haptic device or speaker.

[0197] Memory 540 may be any suitable means of providing storage, such as electrical memory, magnetic memory, or optical memory including RAM, cache, hard disk drive, or removable storage disk. Communication means 560 may include any suitable means of transmitting and receiving signals over a network, such as a network interface chip or device. Components of the computer may be connected in any suitable manner, such as via a physical bus or wirelessly.

[0198] Software 550 (which may be stored in memory 540 and executed by processor 510) may include, for example, programs that embody the functions of this disclosure (e.g., as embodied in the methods described above).

[0199] The software 550 may also be stored and / or transmitted in any non-transitory computer-readable storage medium for use by or in connection with an instruction execution system, device, or apparatus (such as those described above), which may acquire and execute instructions associated with the software from the instruction execution system, device, or apparatus. In the context of this disclosure, the computer-readable storage medium may be any medium (such as memory 1240) that may contain or store a program for use by or in connection with an instruction execution system, device, or apparatus.

[0200] The software 550 can also be propagated within any transmission medium for use by or in connection with an instruction execution system, device, or apparatus (such as those described above), which can acquire and execute instructions associated with the software from the instruction execution system, device, or apparatus. In the context of this disclosure, the transmission medium can be any medium capable of delivering, propagating, or transmitting a program for use by or in connection with an instruction execution system, device, or apparatus. Transmission readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, or infrared wired or wireless propagation media.

[0201] Device 500 can be connected to a network, which can be any suitable type of interconnected communication system. The network can implement any suitable communication protocol and can be protected by any suitable security protocol. The network can include any suitable arrangement of network links that enables the transfer and reception of network signals, such as wireless network connections, T1 or T3 lines, wired networks, DSL, or telephone lines.

[0202] The device 500 can implement any operating system suitable for operation on a network. The software 550 can be written in any suitable programming language, such as C, C++, Java, or Python. In various embodiments, application software embodying the functions of this disclosure can be deployed in different configurations, such as in a client / server setup or via a web browser as a web-based application or web service.

[0203] definition:

[0204] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. Where a definition set forth in this section contradicts or is inconsistent with a definition set forth in a patent, application, publication, or other publication incorporated herein by reference, the definition set forth in this section shall prevail over the definition incorporated herein by reference.

[0205] As used herein, unless the context clearly indicates otherwise, the singular forms “a / an” and “the” include multiple references. Thus, for example, a reference to “peptide” includes one or more peptides, or mixtures of peptides. Furthermore, and unless explicitly stated or obvious from the context, as used herein, the term “or” should be understood as inclusive and covers both “or” and “and”.

[0206] As used herein, the term "about" refers to a common range of error for a corresponding value that is readily known to those skilled in the art. References to "about" values ​​or parameters herein include (and describe) embodiments relating to that value or parameter itself. For example, a description of "about X" includes a description of "X".

[0207] As used herein, the term "analyte" refers to a substance whose chemical composition is being identified and / or measured. In a preferred embodiment, "analyte" refers to a "peptide analyte," and at least a portion of the amino acid sequence and / or identity of the peptide analyte is determined by the methods disclosed herein. In some embodiments, one, two, or more amino acid residues of the peptide analyte are each determined independently with a certain probability, which may be sufficient to determine the identity of the peptide analyte. The peptide analyte is a substrate for the conjugates disclosed herein.

[0208] As used herein, the term "sample" refers to any substance that may contain an analyte for which an analyte assay is required. As used herein, a "sample" can be a solution, suspension, liquid, powder, paste, aqueous, non-aqueous, or any combination thereof. A sample can be a biological sample, such as a biological fluid or biological tissue. Examples of biological fluids include urine, blood, plasma, serum, saliva, semen, feces, sputum, cerebrospinal fluid, tears, mucus, amniotic fluid, etc.

[0209] As used herein, the term "macromolecule" encompasses macromolecules composed of smaller subunits. Examples of macromolecules include, but are not limited to, peptides, polypeptides, nucleic acids, carbohydrates, lipids, macrocyclic compounds, or combinations or complexes thereof.

[0210] As used herein, the term "peptide" encompasses peptides, polypeptides, and proteins, and refers to a molecule comprising a chain of three or more amino acids linked by peptide bonds. In some embodiments, a polypeptide comprises 3 to 50 amino acid residues and is used interchangeably with the term "peptide." In some embodiments, the term "peptide" refers to a polypeptide fragment. In some embodiments, a peptide does not include secondary, tertiary, or higher-order structures. In some embodiments, a polypeptide is a protein. In some embodiments, a protein comprises more than 50 amino acid residues and has secondary, tertiary, or higher-order structures in addition to its primary structure. In some embodiments, a polypeptide is formed by fragmenting a protein with a protease or by other means. The amino acids of a polypeptide are most typically L-amino acids, but may also be D-amino acids, modified amino acids, amino acid analogs, amino acid mimics, or any combination thereof. Polypeptides may be naturally occurring, synthesized, or recombinantly expressed, or produced by a combination of these methods. Polypeptides may also include additional groups that modify the amino acid chain, such as functional groups added via post-translational modification. Polymers may be linear or branched, may include modified amino acids, and may be interrupted by non-amino acid groups. The term also covers amino acid polymers that have been modified naturally or through intervention; for example, disulfide bond formation, glycosylation, esterification, acetylation, phosphorylation, or any other operation or modification, such as conjugation with a detectable tag.

[0211] As used herein, the term "amino acid" refers to an organic compound that serves as a monomeric subunit of a peptide. Amino acids include 20 standard, naturally occurring, or canonical amino acids, as well as non-standard amino acids. Standard, naturally occurring amino acids include alanine (A or Ala), cysteine ​​(C or Cys), aspartic acid (D or Asp), glutamic acid (E or Glu), phenylalanine (F or Phe), glycine (G or Gly), histidine (H or His), isoleucine (I or Ile), lysine (K or Lys), leucine (L or Leu), methionine (M or Met), asparagine (N or Asn), proline (P or Pro), glutamine (Q or Gln), arginine (R or Arg), serine (S or Ser), threonine (T or Thr), valine (V or Val), tryptophan (W or Trp), and tyrosine (Y or Tyr). Amino acids can be L-amino acids or D-amino acids. Non-standard amino acids can be modified amino acids, amino acid analogs, amino acid mimics, non-standard protein amino acids, or naturally occurring or chemically synthesized non-protein amino acids.

[0212] As used herein, the terms "binding agent" or "binding molecule" refer to a nucleic acid molecule, peptide, polypeptide, carbohydrate, or small molecule that binds to, associates with, co-opts with, recognizes, or combines with a binding target, such as a polypeptide or a component or feature of a polypeptide (e.g., a modified terminal amino acid residue). A binding agent may form a covalent or non-covalent association with a polypeptide or a component or feature of a polypeptide. A binding agent may also be a chimeric binding agent composed of two or more types of molecules, such as a nucleic acid-peptide chimeric binding agent or a carbohydrate-peptide chimeric binding agent. A binding agent may be a naturally occurring, synthetically produced, or recombinantly expressed molecule. A binding agent may bind to a single monomer or subunit of a polypeptide (e.g., a single amino acid of the polypeptide) or to multiple linked subunits of a polypeptide (e.g., a longer peptide, polypeptide, or dipeptide, tripeptide, or higher-order peptide within a polypeptide molecule). A binding agent may bind to a linear molecule or a molecule having a three-dimensional structure (also called a conformation). A binding agent may bind to an N-terminal peptide, C-terminal peptide, or intercalated peptide of a peptide, polypeptide, or polypeptide molecule. The conjugating agent can bind to the N-terminal, C-terminal, or intercalated amino acids of a peptide molecule. The conjugating agent preferably binds to chemically modified or labeled amino acids (e.g., amino acids that have been chemically labeled) rather than unmodified or unlabeled amino acids. For example, the conjugating agent may preferably bind to labeled or modified amino acids rather than unlabeled or unmodified amino acids. The conjugating agent can bind to post-translational modifications of the peptide molecule. The conjugating agent can exhibit selective binding to components or features of the peptide (e.g., the conjugating agent can selectively bind to one of 20 possible native amino acid residues and bind with very low affinity or not at all to the other 19 native amino acid residues). The conjugating agent can exhibit lower selective binding, wherein the conjugating agent is capable of binding to or configured to bind to multiple components or features of the peptide (e.g., the conjugating agent can bind to two or more different amino acid residues with similar affinity). The conjugating agent may include an encoded tag that can be linked to the conjugating agent via a linker.

[0213] As used herein, the term "specific binding" refers to a binding reaction between an engineered conjugate and an associated peptide (e.g., a peptide having a specific NTAA residue bound by the conjugate) or a portion thereof, which is more likely to occur than a similar reaction between an engineered conjugate and a random, unassociated peptide. The term "specific" is used herein to indicate the relative affinity of an engineered conjugate to an associated (e.g., a peptide suitable for binding based on designed affinity) peptide. Specific binding generally means that an engineered conjugate is at least twice as likely to bind to an associated peptide as to a random, unassociated peptide (a 2:1 ratio of specific to nonspecific binding). Specific binding to a specific modified NTAA residue of a peptide means that the conjugate binds to the modified NTAA residue with a higher affinity than to the same but unmodified NTAA residue, and to other (structurally different) modified NTAA residues (the modification of the NTAA residue increases the binding affinity between the conjugate and the peptide). In some embodiments, specific binding is not strictly selective, such as the conjugate specifically binding to two or more different modified NTAA residues compared to other modified NTAA residues. For example, the binder may specifically bind to both D- and E-modified NTAA residues compared to other modified NTAA residues (bispecificity). In another example, the binder may specifically bind to V, I, and L-modified NTAA residues compared to other modified NTAA residues (multispecificity). Nonspecific binding refers to background binding and is the amount of signal generated in a binding assay between the engineered binder and the N-terminal modified peptide when the modified NTAA residue associated with the engineered binder is not present at the N-terminus of the peptide. In some embodiments, specific binding refers to the binding between an engineered metalloprotein binder and an N-terminal modified target peptide with a dissociation constant (Kd) of 500 nM or less.

[0214] In some embodiments, the binding specificity between the engineered conjugate and the N-terminal modified target peptide is determined primarily or substantially by the interaction between the engineered conjugate and the modified NTAA residues of the N-terminal modified target peptide, meaning that there is minimal or no interaction between the engineered conjugate and the penultimate terminal amino acid residue (P2) of the target peptide, as well as other residues of the target peptide. In some embodiments, the engineered conjugate binds to the modified NTAA residues of the target peptide with a binding affinity at least 5 times greater than that with any other region of the target peptide. In some embodiments, the engineered conjugate has a substrate-binding pocket having a size and / or geometry that matches the size and / or geometry of the modified NTAA residues of the N-terminal modified target peptide that the engineered conjugate specifically binds to. In such embodiments, the modified NTAA residues occupy the volume of the substrate-binding pocket surrounding the engineered conjugate, effectively preventing the P2 residues of the target peptide from entering the substrate-binding pocket or interacting with the affinity-determining residues of the engineered conjugate. In some embodiments, the engineered conjugate specifically binds to an N-terminal modified target peptide, wherein the target peptide shares the same modified NTAA residues that interact with the engineered conjugate, but has different P2 residues. In some embodiments, the engineered conjugate is capable of specifically binding to each N-terminal modified target peptide from a plurality of N-terminal modified target peptides, wherein the plurality of N-terminal modified target peptides contain at least 3, at least 5, or at least 10 N-terminal modified target peptides, which are modified with the same N-terminal modifier, have the same modified NTAA residues, and have different P2 residues. Therefore, in a preferred embodiment, the engineered conjugate has binding affinity for the modified NTAA residues of the N-terminal modified target peptide, but has very little or low affinity for the P2 or other residues of the target peptide.

[0215] As used herein, the term "selectivity" refers to the ability of a conjugate to preferentially bind to one or more amino acid residues (optionally modified with chemical modifications) of a peptide analyte. In preferred embodiments, "selectivity" describes the ability of a conjugate to preferentially bind to a single N-terminal amino acid residue, or to a small group (e.g., structurally related) of NTAA residues, which are optionally covalently modified with modifications or labels. In some embodiments, a conjugate may exhibit selective binding to specific amino acid residues. In some embodiments, a conjugate may exhibit selective binding to a specific class or type of amino acid residues. In some embodiments, a conjugate may exhibit specific binding kinetics (e.g., a higher binding rate constant and / or a lower dissociation rate constant) to a specific class or type of amino acid residue or modified amino acid residue compared to other amino acid residues or modified amino acid residues. In some embodiments, a conjugate may exhibit selective binding to components of the peptide analyte. In other embodiments, a conjugate may exhibit lower selective binding, wherein the conjugate is capable of binding to or configured to bind to multiple components of the peptide analyte (e.g., the conjugate may bind to two or more different NTAA residues with similar affinity). In some embodiments, the selectivity of each binder to the NTAA residues of the peptide analyte is determined in advance prior to the contact step of the disclosed method.

[0216] As used herein, the terms “encoding assay” and “reverse translation assay” are used interchangeably to refer to a process of encoding a peptide into a nucleic acid sequence by performing a cyclic process comprising: contacting an immobilized peptide with conjugates (independently or as a mixture) from a set of conjugates, and, after binding, extending a recording tag associated with each peptide by transferring nucleic acid information corresponding to the binding event. Each cycle of the encoding assay (“encoding cycle”) results in the extension of the recording tag of the peptide analyte. Typically, the encoding assay has 3 to 20 encoding cycles. Non-limiting examples of encoding assays are disclosed in US 11513126 B2, US 11782062 B2, US 2019 / 0145982 A1, US2023 / 0136966 A1, US 2024 / 0294981 A1 and US 2024 / 0053350 A1, each of which is incorporated herein by reference.

[0217] As used herein, the term "connector" refers to one or more of a nucleotide, nucleotide analogue, amino acid, peptide, polypeptide, polymer, or non-nucleotide chemical motif used to link two molecules. Connectors can be used to link conjugates to coding tags, recording tags to polypeptides, polypeptides to supports, recording tags to solid supports, etc. In some embodiments, connectors link two molecules via enzymatic or chemical reactions (e.g., click chemistry).

[0218] As used herein, the term "barcode" refers to a nucleic acid molecule of about 3 to about 30 bases (e.g., 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30 bases) that provides a unique identifier tag or source information for a polypeptide, a conjugate, a set of conjugates from a single coding cycle (when the set changes between cycles), a sample polypeptide, a set of samples, a polypeptide within a compartment (e.g., a droplet, a bead, or a separate location), a polypeptide within a set of compartments, a portion of a polypeptide, a spatial region or a set of spatial regions, or a library of polypeptides. Barcodes can be artificial sequences or naturally occurring sequences. In some embodiments, each barcode within a group of barcodes is different. In other embodiments, portions of the barcodes in a group of barcodes are distinct, for example, at least about 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 97%, or 99% of the barcodes in the group are distinct. The group of barcodes can be randomly generated or non-randomly generated. In some embodiments, the group of barcodes is error-corrected or fault-tolerant barcodes. Hamming distance, Lie distance, asymmetric Lie distance, Reed-Solomon, Levenshtein-Tenengolts, or similar error-correction methods can be employed. Barcodes can be used to computationally deconvolve multiplexing data and identify sequence reads originating from individual peptides, samples, libraries, etc. Barcodes can also be used to deconvolve collections of peptides assigned to cell blocks for enhancement mapping. For example, instead of mapping peptides back to the proteome, they are mapped back to their source polypeptide molecules or polypeptide complexes. In some embodiments, the term "barcode" also refers to peptide molecules (e.g., peptide barcodes).

[0219] As used herein, the term "encoder barcode" refers to a nucleic acid molecule of about 2 to about 30 bases in length (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30 bases) that provides identification information for its associated binding reagent. The encoder barcode uniquely identifies its associated binding reagent. During the method, the encoder barcode is initially present in an encoding tag attached to the associated binding reagent. During an encoding cycle, when the binding reagent binds to a peptide analyte, the encoder barcode of its complement is transferred to a recording tag associated with the peptide. In some embodiments, the encoder barcode provides identification information for its associated binding reagent and the binding cycle using that reagent. In some embodiments, the encoder barcode may correspond to a binding event that does not occur within the encoding cycle ("empty" encoder barcode sequence). In other embodiments, the encoder barcode is combined with a separate binding cycle-specific barcode within the encoded tag (i.e., a unique sequence for identifying the library of binding reagents used in a particular binding cycle). In some embodiments, the encoder barcode may identify its associated binding reagent as a member of a group of two or more different binding reagents. In some embodiments, the encoder barcode sequence may correspond to two or more binding reagents from a group of binding reagents, and this level of identification is sufficient for the purposes of the analysis. For example, in some embodiments involving binding reagents to amino acids, it may be sufficient to know that the peptide includes one of two possible amino acids at a particular position, rather than definitively identifying the amino acid residue at that position. In other embodiments, where the encoder barcode identifies a group of possible binding reagents, a sequential decoding method may be used to generate a unique identification for each binding reagent. This is accomplished by varying the encoder barcode for a given binding reagent in a repeating cycle of binding (see, Gunderson et al., 2004, Genome Res. 14:870-7). The partial identification coding tag information from each binding cycle, when combined with coding information from other cycles, produces a unique identifier for the binding reagent. For example, a specific combination of coding tags, rather than individual coding tags (or encoder barcodes), provides unique identification information for the binding reagent.

[0220] As used herein, the term “coding tag” or “CT” refers to a polynucleotide of any suitable length, for example, from about 3 bases to about 100 bases (inclusive: any integer including 3 and 100), which includes identification information of its associated conjugate. “Coded tags” can also be made from “sequential polymers” (see, e.g., Niu et al., 2013, Nat. Chem. 5:282-292; Roy et al., 2015, Nat. Commun. 6:7237; Lutz, 2015, Macromolecules 48:4759-4767; each of which is incorporated herein by reference in its entirety). A coding tag may include an encoder sequence (e.g., a barcode including identification information of the conjugate), which optionally has a spacer side-mounted on one side or optionally has a spacer side-mounted on each side. A coding tag may also include an optional UMI and / or an optional coding cycle-specific barcode. Coding tags may be single-stranded or double-stranded. Double-stranded coding tags may include blunt ends, protruding ends, or both. A coding tag may refer to a coding tag directly attached to a conjugate, a complementary sequence that hybridizes with a coding tag directly attached to a conjugate (e.g., for a double-stranded coding tag), or coding tag information present in an extended record tag.

[0221] As used herein, the term “recording tag” or “RT” refers to a portion of, for example, a nucleic acid molecule or a sequenceable polymer molecule (see, e.g., Niu et al., 2013, Nat. Chem. 5:282-292; Roy et al., 2015, Nat. Commun. 6:7237; Lutz, 2015, Macromolecules 48:4759-4767; each incorporated herein by reference in its entirety), from which identification information of the coding tag may be transferred, or identification information about the polypeptide associated with the recording tag (e.g., UMI information) may be transferred from that portion to the coding tag. Identification information may include any information characterizing the molecule, such as information relating to the sample, fraction, partition, spatial location, interacting neighboring molecules, cycle number, etc. Furthermore, the presence of UMI information may also be classified as identification information. In some embodiments, after the conjugate binds to the polypeptide, information from the coding tag linked to the conjugate may be transferred to the recording tag associated with the polypeptide simultaneously with the conjugate binding to the polypeptide. In other embodiments, after the conjugate binds to the polypeptide, information from a recording tag associated with the polypeptide can be transferred to a coding tag attached to the conjugate simultaneously with the binding of the conjugate to the polypeptide. The recording tag can be directly attached to the polypeptide, attached to the polypeptide via a multifunctional linker, or associated with the polypeptide by its proximity (or co-localization) on the support. The recording tag can be attached via its 5' end or 3' end or at an internal site, provided that the attachment is compatible with the method used to transfer coding tag information to the recording tag or vice versa. The recording tag may further include other functional components, such as a universal initiation site, a unique molecular identifier, a barcode (e.g., a sample barcode, a fractionation barcode, a spatial barcode, a compartment tag, etc.), a spacer sequence complementary to the spacer sequence of the coding tag, or any combination thereof. In embodiments where polymerase extension is used to transfer coding tag information to the recording tag, the spacer sequence of the recording tag is preferably located at the 3' end of the recording tag.

[0222] As used herein, the term "spacer" (Sp) refers to a nucleic acid molecule of about 1 to about 20 bases in length (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 bases) located at the end of a recording tag or coding tag. In some embodiments, the spacer sequence is side-mounted to an encoder sequence of the coding tag at one or both ends. After a conjugate binds to a polypeptide, annealing between complementary spacer sequences on its associated coding tag and recording tag, respectively, can transfer binding information to the recording tag, coding tag, or dual-tag construct via primer extension or ligation. Preferably, the spacer sequences within a set of conjugates have the same number of bases. Common (shared or identical) spacers can be used in a set of conjugates. The spacer sequence may have a "cycle-specific" sequence to track conjugates used in a specific coding cycle (i.e., the contact-transfer-release steps of the methods disclosed herein form a "coding cycle"). The spacer sequence (Sp) can be constant across all coding cycles, specific to a particular class of peptide, or specific to a coding cycle number. In some embodiments, only sequential binding of the correct association pairs of RT and CT produces interacting spacer elements and efficient primer extension. The spacer sequence may include a sufficient number of bases to anneal with a complementary spacer sequence in the recording tag to initiate a primer extension (also known as polymerase extension) reaction, or to provide a “snap” for the ligation reaction, or to mediate a “sticky end” ligation reaction. The spacer sequence may include fewer bases than the encoder sequence within the coding tag.

[0223] As used herein, the term "primer extension" (also known as "polymerase extension" and "extension") refers to a reaction catalyzed by a nucleic acid polymerase (e.g., DNA polymerase) in which a nucleic acid molecule annealed to its complementary strand (e.g., an oligonucleotide primer, a spacer sequence) is extended by the nucleic acid polymerase using the complementary strand as a template. Various polymerases capable of performing extensions are known in the art.

[0224] As used herein, the term “unique molecular identifier” or “UMI” refers to a nucleic acid molecule of approximately 3 to approximately 40 bases in length (3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, or 40 bases) that provides a unique identifier tag for each polypeptide, peptide, or conjugate linked to a UMI. Peptide UMIs can be used to computationally deconvolve sequencing data from multiple extended record tags to identify extended record tags originating from individual polypeptides. Peptide UMIs can be used to accurately count the source polypeptide molecules by merging NGS reads into a unique UMI. Conjugate UMIs can be used to identify each individual molecular conjugate bound to a specific polypeptide. For example, UMI can be used to identify the number of independent binding events of a single amino acid-specific binder for a particular peptide molecule. It should be understood that when referring to both UMI and barcode in the context of a binder or peptide, barcode refers to identification information other than the UMI of an independent binder or peptide (e.g., sample barcode, compartment barcode, coded cycle barcode).

[0225] As used herein, the terms “universal priming site” or “universal primer” or “universal priming sequence” refer to nucleic acid molecules that can be used for analysis, amplification, and / or sequencing for extending recording tags. Universal priming sites may include, but are not limited to, priming sites for PCR amplification (primer sequences), flow cell adaptor sequences (which anneal to complementary oligonucleotides on the flow cell surface, enabling bridging amplification on some next-generation sequencing platforms), sequencing priming sites, or combinations thereof. Universal priming sites can be used for other types of amplification, including amplification commonly used in conjunction with next-generation digital sequencing. For example, extending recording tag molecules can be circularized and universal priming sites can be used for rolling circle amplification to form DNA nanospheres that can be used as sequencing templates (Drmanac et al., 2009, Science 327:78-81). Alternatively, recording tag molecules can be circularized and directly sequenced by polymerase extension from universal priming sites (Korlach et al., 2008, Proc.Natl.Acad.Sci.105:1176-1181).

[0226] As used herein, the term "extended recording tag" refers to a recording tag to which information of the coding tag (or its complementary sequence) of at least one binder has been transferred after a conjugate binds to a polypeptide. The information of the coding tag can be transferred directly (e.g., by ligation) or indirectly (e.g., by primer extension) to the recording tag. The information of the coding tag can be enzymatically or chemically transferred (e.g., by chemical ligation) to the recording tag. An extended recording tag may include conjugate information of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 40, 50, or more coding tags. The base sequence of the extended recording tag may reflect the timing and sequential order of binding of the conjugates identified by its coding tag, may reflect a partial sequential order of binding of the conjugates identified by the coding tag, or may not reflect any order of binding of the conjugates identified by the coding tag. In some embodiments, the coded tag information present in the extended recording tag represents the polypeptide sequence being analyzed with at least 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 100% identity. In some embodiments where the extended recording tag does not represent the polypeptide sequence being analyzed with 100% identity, the error may be due to off-target binding of the binder, or due to a “missed” coding cycle (e.g., because the binder failed to bind to the polypeptide during the coding cycle, due to a failed primer extension reaction), or both.

[0227] As used herein, the term "solid support" or "support" refers to any solid material, including porous and non-porous materials, to which peptides can be directly or indirectly associated by any means known in the art, including covalent and non-covalent interactions or any combination thereof. Solid supports can be two-dimensional (e.g., planar surfaces) or three-dimensional (e.g., gel matrices or beads). Solid supports can be any support surface, including but not limited to beads, microbeads, arrays, glass surfaces, silicon surfaces, plastic surfaces, filters, membranes, PTFE membranes, nitrocellulose membranes, nitrocellulose-based polymer surfaces, silicon wafers, flow-through chips, flow cells, biochips including signal transduction electronic devices, channels, microtiter wells, ELISA plates, rotating interferometric disks, nitrocellulose membranes, nitrocellulose-based polymer surfaces, polymer matrices, nanoparticles, microparticles, or microspheres. Materials used for solid supports include, but are not limited to, acrylamide, agarose, cellulose, dextran, nitrocellulose, glass, gold, quartz, polystyrene, polyvinyl acetate, polypropylene, polyester, polymethacrylate, polyacrylate, polyethylene, polyethylene oxide, polysilicate, polycarbonate, polyvinyl alcohol (PVA), Teflon, fluorocarbons, nylon, silicone rubber, polyanhydride, polyglycolic acid, polyvinyl chloride, polylactic acid, polyorthoester, functionalized silanes, polypropyl fumarate, collagen, glycosaminoglycans, polyamino acids, dextran, or any combination thereof. For example, when the solid surface is beads, the beads may include, but are not limited to, ceramic beads, polystyrene beads, polymer beads, polyacrylate beads, methylstyrene beads, agarose beads, cellulose beads, dextran beads, acrylamide beads, solid beads, porous beads, paramagnetic beads, glass beads, controlled-pore beads, silica-based beads, or any combination thereof. Beads may be spherical or irregularly shaped. The size of the beads can range from several nanometers (e.g., 10 nm) to several millimeters (e.g., 1 mm). In some embodiments, the size of the beads ranges from about 0.2 micrometers to about 200 micrometers, or from about 0.5 micrometers to about 5 micrometers. In some embodiments, the term "bead" in solid support can refer to a single bead or multiple beads. In some embodiments, the size of the nanoparticles ranges from about 10 nm to about 500 nm in diameter.

[0228] As used herein, the terms "nucleic acid" or "polynucleotide" refer to single-stranded or double-stranded polynucleotides containing deoxyribonucleotides or ribonucleotides linked by 3'-5' phosphodiester bonds, as well as polynucleotide analogs. Nucleic acid molecules include, but are not limited to, DNA, RNA, and cDNA. Polynucleotide analogs may have a backbone in addition to the standard phosphodiester bonds found in native polynucleotides, and optionally have one or more modified sugar moieties in addition to ribose or deoxyribose. Polynucleotide analogs contain bases capable of hydrogen bonding with standard polynucleotide bases via Watson-Crick base pairing, wherein the analog backbone presents bases in a manner that allows such hydrogen bonding between bases in the oligonucleotide analog molecule and the standard polynucleotide in a sequence-specific manner. Examples of polynucleotide analogs include, but are not limited to, xenobiotic nucleic acids (XNA), bridged nucleic acids (BNA), glycol nucleic acids (GNA), peptide nucleic acids (PNA), γPNA, morpholinopolynucleotides, locked nucleic acids (LNA), threononucleotides (TNA), 2'-O-methyl polynucleotides, 2'-O-alkylribosyl-substituted polynucleotides, phosphate thioester polynucleotides, and borosilicate polynucleotides. Polynucleotide analogs may have purine or pyrimidine analogs, including, for example, 7-denitropurine analogs, 8-halopurine analogs, 5-halopyrimidine analogs, or universal base analogs that can pair with any base, including hypoxanthine, nitroazole, isoquinolone analogs, azolecarbamate, and aromatic triazole analogs, or base analogs with additional functions, such as a biotin moiety for affinity binding. In some embodiments, the nucleic acid molecule or oligonucleotide is a modified oligonucleotide. In some embodiments, the nucleic acid molecule or oligonucleotide is DNA with pseudo-complementary bases. In some embodiments, nucleic acid molecules or oligonucleotides are modified by backbone modification, sugar modification, or nucleobase modification.

[0229] As used herein, “nucleic acid sequencing” means determining the sequence of nucleotides in a nucleic acid molecule or a sample of nucleic acid molecules. It also refers to any possible sequencing method among the many sequencing methods known in the art. Examples of sequencing methods include, but are not limited to, next-generation sequencing, sequencing by synthesis, ligation sequencing, hybridization sequencing, polony sequencing, ion semiconductor sequencing, nanopore sequencing, single-molecule sequencing, and pyrosequencing. Further examples of nucleic acid sequencing technologies include platforms provided by Illumina, BGI, Qiagen, Thermo-Fisher, and Roche, including formats such as parallel bead arrays, sequencing by synthesis, ligation sequencing, capillary electrophoresis, electronic microarrays, “biochips,” microarrays, parallel microarrays, and single-molecule arrays (see, for example, Service, Science (2006) 311:1544-1546). Some sequencing methods rely on amplification to clone many nucleic acid (e.g., DNA) molecules in parallel for sequencing in a staged approach. Single-molecule sequencing detects a single DNA molecule and does not require amplification or synchronization. Examples of single-molecule sequencing methods include real-time single-molecule sequencing (Pacific Biosciences), nanopore-based sequencing (Oxford Nanopore), double-strand break nanopore sequencing, and direct imaging of DNA using advanced microscopy.

[0230] As used herein, “analyzing” a peptide means identifying, detecting, quantifying, characterizing, differentiating, or a combination thereof all or part of the components of a peptide. For example, analyzing a peptide, polypeptide, or polypeptide includes determining all or part of the amino acid sequence (continuous or discontinuous) of the peptide. Analyzing a peptide also includes the partial identification of the components of the peptide (e.g., the NTAAs of the peptide). For example, partial identification of amino acid residues in the peptide sequence can identify amino acids in the peptide as belonging to a subset of possible amino acid residues. In preferred embodiments, peptide analysis typically begins with the analysis of nNTAAs and then continues to the next amino acid residue of the peptide (i.e., n-1, n-2, n-3, etc.). This is accomplished by removing nNTAAs, thereby converting the n-1 amino acid residue of the peptide to an N-terminal amino acid (referred to herein as “n-1NTAA”). Analyzing a peptide may also include determining the presence and frequency of post-translational modifications on the peptide, which may or may not include information about the sequential order of post-translational modifications on the peptide. Analyzing a peptide may also include determining the presence and frequency of epitopes in the peptide, which may or may not include information about the sequential order or location of epitopes within the peptide. Analysis of peptides may include combining different types of analysis, such as obtaining epitope information, amino acid sequence information, post-translational modification information, or any combination thereof.

[0231] As used herein, the term "detectable label" refers to a substance that, when associated with another substance, indicates the presence of that substance. A detectable label may be a substance linked to or incorporated into a substance to be detected. In some embodiments, a detectable label is suitable for both detection and quantification, for example, a detectable label that emits a detectable and measurable signal. Detectable labels include any labels available and compatible with the provided peptide assay format, and include, but are not limited to, bioluminescent labels, biotin / avidin labels, chemiluminescent labels, chromophores, coenzymes, dyes, electroactive groups, electrochemiluminescent labels, enzyme labels (e.g., alkaline phosphatase, luciferase, or horseradish peroxidase), fluorescent labels, latex particles, magnetic particles, metals, metal chelates, phosphorescent dyes, peptide labels, radioactive elements or fractions, and stable free radicals. When attached to a conjugate, a detectable label may indicate a binding event between the conjugate and the peptide analyte.

[0232] The term “unmodified” (also known as “wild-type” or “natural”) used in this article, in conjunction with biological materials such as nucleic acid molecules and peptides, refers to biological materials found in nature that have not been modified by human intervention.

[0233] The terms “modified,” “engineered,” “variant,” or “mutant” (e.g., engineered conjugate) used with respect to nucleic acid and polypeptide molecules refer to molecules produced through human intervention and / or that are not naturally occurring. A variant, mutant, or engineered conjugate is a polypeptide with an altered amino acid sequence relative to an unmodified or wild-type polypeptide (such as a starting scaffold) or a portion thereof. An engineered conjugate is a polypeptide differing from the wild-type scaffold sequence or a portion thereof by one or more amino acid substitutions, deletions, additions, or combinations thereof. The sequence of an engineered conjugate may contain 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 50, 100, or more amino acid differences (e.g., mutations) compared to the sequence of the starting scaffold. Engineered conjugates typically exhibit at least 30%, 40%, 50%, 60%, 70%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99% or more sequence identity with the corresponding wild-type starting scaffold. Both naturally occurring and non-naturally occurring amino acids are included within the permitted range of substitutions or additions. Engineered conjugates are not limited to any engineered conjugates manufactured or generated by a particular manufacturing method, but include engineered conjugates manufactured or generated, for example, by genetic selection, peptide engineering, chemical synthesis, directed evolution, de novo recombinant DNA technology, or combinations thereof. In the context of variants or engineered conjugates, the term "variant" should not be construed as imposing any conditions on any particular starting composition or method for producing a variant or engineered conjugate. Therefore, a variant or engineered conjugate refers to a composition and is not necessarily a product produced by any given process.

[0234] As used herein, the term "modified amino acid residue" refers to an amino acid residue within a polypeptide, including modifications that distinguish it from its corresponding native or unmodified amino acid residue. In some embodiments, the modification may be a post-translational modification of the naturally occurring amino acid residue. In other embodiments, the modification is a modification of the amino acid residue that is not naturally occurring; such modified amino acid residues are not naturally present in peptides of living organisms (representing non-natural amino acid residues). Such modified amino acid residues can be manufactured by modifying native amino acid residues within a polypeptide with a modifying agent (such as an N-terminal modifier), or can be chemically synthesized and incorporated into the peptide during peptide synthesis.

[0235] In some embodiments, variants of engineered peptides exhibiting only non-substantial or negligible structural differences can be generated by making conserved amino acid substitutions within the engineered peptide. By doing so, engineered peptide variants comprising sequences having at least 80% (85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, and 99%) sequence identity with the engineered peptide sequence can be generated, retaining at least one functional activity of the engineered peptide, such as the ability to specifically bind to the N-terminal amino acid (NTAA) residues of the peptide analyte. Examples of conserved amino acid variations are known in the art. Examples of nonconserved amino acid changes that can cause significant alterations to the polypeptide structure include changes that result in: (a) a hydrophilic residue (e.g., serine or threonine) replacing (or being replaced by) a hydrophobic residue (e.g., leucine, isoleucine, phenylalanine, valine, or alanine); (b) a cysteine ​​or proline replacing (or being replaced by) any other residue; (c) a residue with a positively charged side chain (e.g., lysine, arginine, or histidine) replacing (or being replaced by) a negatively charged residue (e.g., glutamic acid or aspartic acid); or (d) a residue with a large side chain (e.g., phenylalanine) replacing (or being replaced by) a residue without a side chain (e.g., glycine). Methods for targeted amino acid substitutions, deletions, truncations, and insertions are well known in the art. For example, amino acid sequence variants can be prepared by mutations in DNA. Methods for polynucleotide alteration are well known in the art, for example, Kunkel et al. (1987) Methods in Enzymol. 154:367-382; U.S. Patent No. 4,873,192 and references cited therein.

[0236] As used herein, “identifying” a peptide means predicting the identity of the peptide with a certain probability. This can be done by identifying the components of the peptide (e.g., one or more amino acid residues). It can also be done by predicting certain amino acid residues of the peptide and their positions with a certain probability, thereby creating a peptide signature profile, and then bioinformatically matching the resulting peptide signature profile with corresponding signature profiles of peptides that may be present in the sample (e.g., by matching the peptide signature profile with peptide sequences from a proteomic or genomic database). For example, in some embodiments, the existing selectivity of the binder is insufficient to definitively determine the NTAA residues that the binder binds to. In these cases, the identity of the NTAA residues can be determined with a certain probability (e.g., it is D, E, or H instead of A, G, I, or L). Subsequent similar determinations of adjacent amino acid residues create an array of possible variants of the peptide based on the variants in the determined amino acid residues, and this array of variants can be narrowed down to specific sequences (if enough amino acid residues have been determined).

[0237] The term "sequence identity" is a measure of the identity between peptides at the amino acid level and between nucleic acids at the nucleotide level. Peptide sequence identity is determined by comparing the amino acid sequence at a given position in each sequence when aligning the sequences. Similarly, nucleic acid sequence identity is determined by comparing the nucleotide sequence at a given position in each sequence when aligning the sequences. "Sequence identity" refers to the percentage of identical subunits at corresponding positions in two sequences when aligning them to maximize subunit matching; that is, taking into account gaps and insertions. For example, the BLAST algorithm (NCBI) calculates the sequence identity percentage and performs statistical analysis on the similarity and identity between two sequences. Software used for BLAST analysis is publicly available on the website of the National Center for Biotechnology Information (NCBI).

[0238] The terms "corresponding to a position within a polypeptide or polynucleotide" or "position relating to a position within a polypeptide or polynucleotide," such as the statement that a nucleotide or amino acid position "corresponds" to a nucleotide or amino acid position in a published sequence (such as the sequence listed in the sequence listing), refer to the nucleotide or amino acid position identified in a polynucleotide or polypeptide when aligned with a published sequence using a standard alignment algorithm (such as the BLAST algorithm (NCBI)). Those skilled in the art can identify any given amino acid residue in a given polypeptide at a position corresponding to a specific position in a reference sequence (such as the position listed in the sequence listing) by aligning the polypeptide sequence with a reference sequence (e.g., by using BLASTP, which is publicly available on the NCBI website).

[0239] The term “connecting” or “attaching” one substance to another means connecting or linking these substances together using one or more covalent bonds and / or non-covalent interactions. Some examples of non-covalent interactions include hydrogen bonds, hydrophobic bonds, and van der Waals forces. The connection can be direct or indirect, such as via a joint or via another part. In a preferred embodiment, linking two or more substances together does not impair the structure or functional activity of the linked substances. The term “associated with” (e.g., one substance is associated with another substance) means bringing two substances together so that they can synergistically participate in the methods described herein. In a preferred embodiment, the association of the two substances preserves their structure and functional activity. The association can be direct or indirect. When one substance is directly associated with another substance, this is equivalent to one substance connecting or attaching to the other substance. An indirect association means that two substances are brought together by means other than direct connection or attachment. In some embodiments, an indirect association means that the two substances are co-located with each other or located close to each other.

[0240] As used herein, the term "macromolecule including a component" means that the component is part of the macromolecule or is directly attached to the macromolecule via one or more covalent bonds that bind them together as a single molecule. Conversely, the term "macromolecule associated with a component" indicates that the component may or may not be directly attached to the macromolecule via one or more covalent bonds, but may be associated or colocalized with the macromolecule through non-covalent interactions, or alternatively, indirectly associated via a solid support (e.g., when the macromolecule is attached to the solid support, and the component is attached to the solid support in close proximity to the macromolecule). For example, "macromolecule associated with a recording tag" covers the various possible ways in which a macromolecule and a recording tag are associated (direct, covalent, or non-covalent association, or indirect association, such as via a connector or via another object, such as via a solid support). The terms "attachment" and "connection" are used interchangeably and refer to covalent or non-covalent attachment.

[0241] As used in this article, the term "peptide bond" refers to a chemical bond formed between two molecules (such as two amino acids) when the carboxyl group of one molecule reacts with the amino group of another molecule (releasing water molecules (H2O)).

[0242] The term "binding character" (also known as the binding-specific character of a binder) conveys information about the binding and / or encoding activity of a binder in an assay encoding two or more peptide analytes. In some embodiments, the binding character refers to the binding and / or encoding activity of a given concentration of a binder for various NTAA-modified peptides at their last and penultimate amino acid residues (e.g., representative exemplary binding characteristics in...). Figures 16A to 16G(As given in the text). In these examples, the binding signature of the conjugate to the P1-P2 residues of the immobilized peptide conveys information about the extent to which the conjugate binds to and encodes different modified NTAA residues of the peptide (the binding signature depends on the P2 residue of the peptide). In other embodiments, the binding signature conveys information about the binding and / or encoding activity of a single conjugate for the entire polypeptide sequence in a sample at a given concentration in a coding assay. The coding activity is characterized by the coding yield (or coding rate), which is the fraction of recording tags encoded at each cycle (the fraction of recording tags extended by encoder barcodes transferred from the coding tags of the conjugate in a given coding cycle). The coding yield of the conjugate depends on the conjugate concentration in the coding assay. The coding yield of the conjugate can also change when coding assay parameters are changed, such as conjugate incubation time, ligase incubation time, buffer conditions, washing steps, etc.

[0243] The binding signature of a conjugate can change when it mixes with other conjugates that can be substrates and compete with it (such as for modified NTAA residues of a peptide). In a given set of conjugates used in coding assays, the binding signature of each conjugate in that set can be determined from its coding activity, even in the presence of other conjugates (see [link to assay]). Figures 16A to 16GWhen the concentration of binders in a group of binders changes, the binding characteristics of the binders can change. In methods for inferring the amino acid sequence of a peptide from a binder identifier string generated in a coding assay, the binding characteristics of each binder in a group of binders used in the coding assay can be determined in advance (from a previously generated coding dataset obtained using substantially the same coding assay parameters at one or more different binder concentrations), or can be determined simultaneously during other methodological steps (e.g., by including a baseline truth sample within the peptide analyte). In some embodiments, when a previously generated coding dataset or actual coding data is input into a model, a computer model can calculate (determine) the binding characteristics of each binder in a group of binders used in the coding assay, thereby enabling the model to infer binder specificity and coding activity. In some embodiments, the binding characteristics of each binder can be implicitly determined from the coding activity of the binder using a training dataset with known peptide sequences, such as publicly available algorithms that internally learn a representation of the binding characteristics of a group of binders. In some embodiments, the coding activity of the binder is implicit in the underlying dataset and can be determined by training a machine learning model on the dataset, and although the binding characteristics may not be directly analyzed, binding characteristic information can be captured and learned by updating the parameters of the machine learning model during training. In a non-restrictive instance, in a trained Naive Bayes model approach (see Example 10), the conditional probability of a given amino acid identity at index i in a polypeptide sequence co-occurring with a binder label at index j is calculated. Therefore, implicitly within the trained model is a learned binding feature for each binder (denoted as a binder label) with respect to each amino acid identity (denoted as an amino acid label) at a given position in the polypeptide, such that the model learns binding features from the training dataset, even though no supplementary binding feature information from other encoded assay datasets is explicitly input into the computer model during training. Instead, binding features are learned from the same underlying dataset during the training process, and the implicit representation of the learned binding features can then be used to decode polypeptide sequences on the test dataset or supplementary datasets.

[0244] The term "probability score" generally refers to a value representing the likelihood that an entity is classified into a particular category. In a preferred embodiment, the probability score is between 0 and 100%. In some embodiments, "probability score" may refer to empirical probability (also known as experimental probability) or theoretical probability and its derivatives, including logit, log probability, likelihood, log-likelihood, odds, log odds, frequency, maximum likelihood estimate and / or any metric that can be statistically inferred to represent the probability or expectation of a correct association between one or more binding identifiers and one or more amino acid sequences of a polypeptide, as disclosed in the methods herein. In some embodiments, the disclosed machine learning model may generate a probability score in log odds form, which may be transformed into a probability using a softmax function. Conversely, probabilities in polypeptide decoding algorithms may be transformed into logit using a logit function. In some embodiments, when decoding a binding identifier string to infer a polypeptide sequence, any metric that can be statistically inferred and transformed into a probability (the probability that a given polypeptide sequence is inferred from or associated with a given binding identifier string) is encompassed in the meaning of "probability score" that may be generated by the decoding algorithms described herein.

[0245] A “numerical representation” (also called an “embedding”) of a conjugate identifier string, polypeptide sequence, or other amino acid sequence of a conjugate, lyase, or ligase is defined as an N-dimensional tensor (where N is a positive rational number) containing a real number at each element of the tensor. In some embodiments, information in the input sequence string (e.g., containing alphanumeric characters or amino acid identity) is transformed into a tensor of real numbers, where the information in the sequence resides in a latent space that can be further transformed by a function and / or input into a machine learning model, deep learning model, or artificial intelligence model.

[0246] Throughout this disclosure, all aspects of the invention are presented in a scope format. It should be understood that the use of a scope format is for convenience and brevity only and should not be construed as a rigid limitation on the scope of the invention. Therefore, the scope description should be considered as having specifically disclosed all possible sub-scopes and the individual values ​​within those scopes. For example, a description of a scope such as 1 to 6 should be considered as having specifically disclosed sub-scopes such as 1 to 3, 1 to 4, 1 to 5, 2 to 4, 2 to 6, 3 to 6, etc., and individual numbers within that scope, such as 1, 2, 3, 4, 5, and 6. This applies regardless of the breadth of the scope.

[0247] The use of sequential terms (such as "first," "second," "third," etc.) to modify claim elements in claims does not in itself imply any priority, precedence, or order of one claim element relative to another, or the chronological order of actions of the method. Rather, they serve merely as markers to distinguish one claim element with a certain name from another element with the same name (if sequential terms were not used). Similarly, the use of a), b), etc., or i), ii), etc., does not in itself indicate any priority, precedence, or order of steps in the claims. Likewise, the use of these terms in the specification does not in itself indicate any desired priority, precedence, or order.

[0248] Other objects, advantages and features of the invention will become apparent from the following description taken in conjunction with the accompanying drawings.

[0249] Description of exemplary embodiments

[0250] This paper provides a computer-implemented method for analyzing multiple peptides encoded in multiple nucleic acid sequences using a coding assay. The method includes: (a) receiving, at one or more processors, multiple nucleic acid sequences generated from a coding assay, wherein each nucleic acid sequence comprises a series of encoder barcode sequences, and wherein each encoder barcode sequence in the given series corresponds to a conjugate from a set of conjugates that bind to one or more components of the peptide in the coding assay; and (b) generating a conjugate identifier string for each nucleic acid sequence in the multiple nucleic acid sequences based on the corresponding series of encoder barcode sequences, thereby generating multiple peptides corresponding to the multiple nucleic acid sequences. (c) Using one or more processors, inferring amino acid sequences of peptides from multiple peptides from the binder identifier strings, the inference being based on (i) binding characteristics of binders from a set of binders corresponding to binder identifiers in the binder identifier strings, and (ii) a calculated probability score of the association between one or more binder identifiers in the binder identifier strings and one or more amino acid sequences of peptides in multiple peptides; and (d) based on the calculated probability score and the inferred amino acid sequence, outputting data relating to at least partial identity and / or quantity of at least one peptide in multiple peptides.

[0251] This document also provides a computer-implemented method for analyzing multiple peptides encoded in multiple nucleic acid sequences using a coding assay, the method comprising: a) receiving, at one or more processors, multiple nucleic acid sequences generated from a coding assay, wherein each nucleic acid sequence comprises a series of encoder barcode sequences, and wherein each encoder barcode sequence in the given series of encoder barcode sequences corresponds to a conjugate that binds to a component of a peptide in the multiple peptides in the coding assay; b) using one or more processors to assign conjugate identifiers to each encoder barcode sequence in the series of encoder barcode sequences in each of the multiple nucleic acid sequences to generate multiple conjugate identifier strings for the multiple nucleic acid sequences; c) for each peptide and / or peptide fragment in the multiple peptides, generating in a computer a set of simulated conjugate identifiers based on predetermined parameters of the coding assay. The identifier string, wherein predetermined parameters include: i) the probability of correctly or incorrectly assigning one or more conjugate identifiers to at least one component of a given polypeptide or polypeptide fragment based on the binding characteristics of conjugates used in the coding assay; and ii) optionally, for at least one component of a given polypeptide or polypeptide fragment, the probability of successfully cleaving the at least one component after a conjugate binds to the at least one component in the coding assay, thereby determining the probability of each simulated conjugate identifier string generated from the given polypeptide or polypeptide fragment; d) for each of the plurality of conjugate identifier strings generated in (b), calculating the probability of the given conjugate identifier string being generated from one or more polypeptides or fragments thereof based on the probability of the simulated conjugate identifier strings determined in (c), thereby determining at least partial identity and / or quantity of at least one polypeptide among a plurality of polypeptides.

[0252] This document also provides a computer-implemented method comprising: a) receiving amino acid sequences of multiple peptides at one or more processors; b) generating multiple peptide sequences based on the amino acid sequences using the one or more processors; c) generating multiple peptide reads based on the multiple peptide sequences using the one or more processors; d) converting each of the multiple peptide reads into an identifier string using the one or more processors based on a sequence of multiple binders and determined binding characteristics, wherein each of the multiple binders binds to a component of the peptide and corresponds to a binder identifier, and wherein the binder identifier corresponds to a barcode sequence; e) providing the multiple identifier strings as input to a trained model using the one or more processors, wherein the trained model is configured to convert the identifier strings into nucleic acid sequences and output multiple nucleic acid sequences corresponding to multiple peptides, the conversion being based on a set of probabilities of multiple binders binding to components of the peptides in the multiple peptides.

[0253] Various embodiments are equally applicable to the aspects provided herein, but will be described only once for the sake of brevity. Therefore, the various embodiments described below are also applicable to the aspects described below.

[0254] In some embodiments of the disclosed method, the components of the polypeptide bound by the binder include polypeptides or portions thereof obtained by fragmenting polypeptides from a variety of polypeptides.

[0255] In some embodiments of the disclosed method, the components of the polypeptide bound by the binder include one or more amino acid residues.

[0256] In some embodiments of the disclosed method, the components of the polypeptide bound by the binder include post-translational modifications.

[0257] In some embodiments of the disclosed method, the components of the polypeptide bound by the binder include modified amino acid residues.

[0258] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide includes: (i) for each of a plurality of binder identifier strings, converting a given binder identifier string into one or more peptide reads, the conversion being based on the binding characteristics of binders in a set of binders corresponding to binder identifiers present in the given binder identifier string; and (ii) calculating a probability score for each of the one or more peptide reads, wherein the probability score indicates the probability that a given peptide read generates a given binder identifier string.

[0259] In some embodiments, the disclosed method further includes: for each of a plurality of binder identifier strings, filtering out peptide reads generated for one or more peptide reads for a given binder identifier string, the filtering being based on (i) a probability score for each peptide read, and / or (ii) the probability that a given peptide read is generated from an amino acid sequence of a plurality of polypeptides.

[0260] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide includes: (i) generating one or more simulated binder identifier strings for each amino acid sequence of a plurality of polypeptides using one or more parameters of a coding assay; (ii) calculating a probability score for each of the one or more simulated binder identifier strings, or for each of the one or more amino acid sequences of a polypeptide in a plurality of polypeptides, based on the probability that a given simulated binder identifier string is associated with one or more amino acid sequences of a polypeptide in a plurality of polypeptides; and (iii) matching each of the plurality of binder identifier strings with one or more simulated binder identifier strings based on the calculated probability scores of the one or more simulated binder identifier strings.

[0261] In some embodiments of the disclosed method, one or more parameters of the encoding assay include the efficiency of functionalization of N-terminal amino acid (NTAA) residues of peptides in multiple peptides, the efficiency of cleavage of NTAA residues of peptides in multiple peptides, the efficiency of encoding of NTAA residues of peptides in multiple peptides, or any combination thereof.

[0262] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide includes: inputting a binder identifier string into a trained machine learning model, wherein the trained machine learning model is trained on empirically determined coding assay data.

[0263] In some embodiments of the disclosed method, the trained machine learning model is trained using a training dataset comprising one or more isolated peptide samples or their digital representations of binder identifier string data or their digital representations, the isolated peptide samples undergoing a coding assay using the same sample preparation protocol as that used for processing multiple peptides.

[0264] In some embodiments of the disclosed method, the trained machine learning model is configured to (i) map each of a plurality of binder identifier strings to a specific polypeptide sequence, or (ii) assign a given binder identifier string fractionally to two or more specific polypeptide sequences as part of inferring the amino acid sequence of the polypeptide from the binder identifier.

[0265] In some embodiments of the disclosed method, the assay parameter data determined empirically includes: (i) the probability of assigning one or more binder identifiers from a set of binders to potential N-terminal amino acid (NTAA) residues of the polypeptide based on the binding characteristics of the binders used in the coding assay; and (ii) the probability of successfully cleaving the N-terminal amino acid residues after encoding them in the coding assay for each potential N-terminal amino acid residue in the polypeptide.

[0266] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide from a binder identifier string includes: inputting the binder identifier string or its numerical representation into a trained machine learning model, wherein the trained machine learning model is trained on a training dataset comprising a set of simulated binder identifier strings or their numerical representations generated for a given input distribution of the polypeptide, the generation being based on assay parameter data determined empirically, the assay parameter data including:

[0267] (i) The probability of assigning one or more binder identifiers from a set of binders to potential N-terminal amino acid (NTAA) residues of a polypeptide based on the binding characteristics of the binders used in the coding assay.

[0268] (ii) For each potential NTAA residue in the polypeptide, the probability of cleaving the NTAA residue after the NTAA is encoded in the coding assay, and (iii) the probability of modifying the NTAA residue.

[0269] In some embodiments of the disclosed method, a set of binders includes at least five different binders, and each binder binds one or more components of a polypeptide that each includes NTAA residues of a polypeptide from a variety of polypeptides or NTAA residues modified by an N-terminal modifier.

[0270] In some embodiments of the disclosed method, a set of binders includes at least five different binders, and each binder binds one or more components of a polypeptide that each includes a terminal amino acid or a terminal dipeptide of a polypeptide from a plurality of polypeptides.

[0271] In some embodiments of the disclosed method, the trained machine learning model is trained on a training dataset generated by performing multiple coding assays with a predetermined analyte and by inputting a series of encoder barcode sequences or their digital representations and coding assay parameters generated during each coding assay into the machine learning model.

[0272] In some embodiments of the disclosed method, the computer model infers two or more amino acid sequences of a polypeptide from a plurality of polypeptides from each binder identifier string and outputs the probability that a given binder identifier string in the plurality of binder identifier strings originates from one of the two or more amino acid sequences of the polypeptide inferred from the binder identifier string.

[0273] In some embodiments of the disclosed method, (i) each of the plurality of nucleic acid sequences further includes one or more auxiliary sequences; (ii) the method further includes using the one or more processors to assign auxiliary identifiers to at least one of the one or more auxiliary sequences in each of the plurality of nucleic acid sequences to generate a plurality of binder identifier strings including the auxiliary identifiers; and (iii) the one or more auxiliary sequences include one or more identifier sequences or their complements, one or more spacer sequences or their complements, one or more sequencing primer sequences or their complements, or any combination thereof. In some embodiments, the one or more identifier sequences include one or more sample barcode sequences, one or more bead barcode sequences, one or more unique molecular identifier (UMI) sequences, or any combination thereof.

[0274] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide from a binder identifier string includes inputting the binder identifier string and coding assay data into a computer model. In some embodiments, the computer model is configured to identify a unique binder identifier feature profile among a plurality of binder identifier strings as part of the inferred amino acid sequence of the polypeptide from the binder identifier, and wherein a given unique binder identifier feature profile includes a set of binder identifier strings associated with a single polypeptide among a plurality of polypeptides. In some embodiments, the computer model calculates the binding characteristics of binders in a set of binders based on the input coding assay parameter data.

[0275] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide from a binder identifier includes inputting one or more corresponding numeric representations of the binder identifier string and optionally one or more corresponding numeric representations encoding assay parameter data into a computer model.

[0276] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide from a binder identifier includes outputting one or more corresponding numerical representations of the polypeptide sequence from a computer model. In some embodiments, the corresponding numerical representation of the binder identifier string is a binder sequence embedding generated from a trained protein language model (pLM), protein basic model, or natural language processing (NLP) model. In some embodiments, the corresponding numerical representation of the polypeptide sequence is a polypeptide sequence embedding generated from a protein language model, protein basic model, or natural language processing model.

[0277] In some embodiments of the disclosed method, the corresponding numerical representations of the encoded assay parameter data include: (i) the concentration of one or more conjugates in a set of conjugates; (ii) the thermodynamic parameters of the conjugates, including the binding rate constant and dissociation rate constant of one or more polypeptide components; (iii) the incubation time of the conjugates; (iv) the washing time of the conjugates after incubation; (v) the ligase concentration; (v) the ligase reaction time; (vi) the thermodynamic parameters of the ligase, including the Michaelis constant and catalytic turnover; (vii) the estimated polypeptide concentration; (viii) the concentration of the lyase; and (ix) the buffer conditions for each step of the encoded assay, including the enzyme substrate concentration, salt type, salt concentration, and pH.

[0278] In some embodiments of the disclosed method, inferring the amino acid sequence of a polypeptide from binder identifiers in a plurality of binder identifier strings includes: (c1) converting the plurality of binder identifier strings into a plurality of peptide reads based on determined binding characteristics of the binders, the binders corresponding to binder identifiers assigned in a given binder identifier string; and (c2) providing the plurality of peptide reads as input to a computer model using one or more processors, wherein the computer model is configured to assign the peptide reads to the amino acid sequences of polypeptides in a plurality of polypeptides and output data relating to at least partial identity and / or quantity of at least one polypeptide in the plurality of polypeptides.

[0279] In some embodiments of the disclosed method, the output of the computer model includes: (i) partial identity and / or quantity of at least one of a plurality of polypeptides, or (ii) data from which a user derives partial identity and / or quantity of at least one of a plurality of polypeptides.

[0280] In some embodiments of the disclosed method, for a nucleic acid sequence among a plurality of nucleic acid sequences generated from a coding assay and received at one or more processors, the output of the computer model includes one or more amino acid sequences of a polypeptide among a plurality of peptides. In some embodiments of the disclosed method, for a given nucleic acid sequence among a plurality of nucleic acid sequences generated from a coding assay, the computer model outputs several amino acid sequences and optionally calculates a probability score for each of the plurality of amino acid sequences. These probability scores can be used to predict which polypeptide is present in the analyzed sample. In some embodiments of the disclosed method, the computer model outputs one or more amino acid sequences providing at least partial identity to a polypeptide among a plurality of peptides. The term “partial identity” of a polypeptide refers to an amino acid sequence obtained by the decoding method that is homologous or identical to a portion of the polypeptide amino acid sequence, and from which the entire polypeptide amino acid sequence can be inferred with a certain probability. For example, for a given nucleic acid sequence among a plurality of nucleic acid sequences generated from a coding assay, the disclosed computer model may output several amino acid sequences, and one or more of these amino acid sequences may each have at least 50%, at least 60%, at least 70%, at least 80%, or at least 90% identity to a portion of one or more amino acid sequences of a polypeptide among a plurality of peptides. For example, given a nucleic acid sequence, the output amino acid sequence may correspond to one or more homologs of a polypeptide from a variety of peptides. In some embodiments of the disclosed method, the computer model outputs an "imperfect" or "partial" amino acid sequence, which is an amino acid sequence in which one or more amino acid residues are undefined (i.e., labeled Z, which can be any amino acid residue). Imperfect or partial amino acid sequences can still be used to identify polypeptides from a variety of peptides with a certain degree of confidence.

[0281] In some embodiments of the disclosed method, the output of the computer model includes data relating to the relative amounts of one or more peptides among a plurality of peptides. In some embodiments, the relative amounts of one or more peptides may be calculated or estimated based on the amount of a specific binder identifier string (or a nucleic acid sequence comprising a specific set of encoder barcode sequences) generated from a coding assay.

[0282] The binder identifier within a given binder identifier string can be a nucleic acid sequence, an abstract binder identifier (e.g., a binder name), or another object corresponding to an encoder barcode sequence within a given nucleic acid sequence of a plurality of nucleic acid sequences generated from an encoding assay and received at one or more processors. In some embodiments of the disclosed method, each binder identifier is assigned to a specific encoder barcode sequence within a nucleic acid sequence of the plurality of nucleic acid sequences by inferring identification information about the corresponding binder of the encoder barcode sequence. In some embodiments of the disclosed method, the assigned binder identifiers for each nucleic acid sequence of the plurality of nucleic acid sequences are then organized into a series of binder identifiers (i.e., binder identifier strings) that reflect the identity and order of binders used during encoding a specific peptide in the encoding assay. In other words, each of the plurality of binder identifier strings includes information about the identity and order of binders used during encoding a specific peptide of a plurality of peptides in the encoding assay. The plurality of binder identifier strings include information about the identity and order of binders used in the encoding assay to encode multiple peptides into multiple nucleic acid sequences.

[0283] In some embodiments of the disclosed method, a binder identifier string is generated for each of a plurality of nucleic acid sequences generated from a coding assay, based on a corresponding series of encoder barcode sequences. In some embodiments, the computer model generates the binder identifier string in advance before starting the inference step of the disclosed method. In other embodiments, each of the plurality of binder identifier strings is generated concurrently with the inference step of the disclosed method. In these embodiments, the computer model begins the inference step by analyzing a single binder identifier, then considers the next binder identifier detected in a given series of encoder barcode sequences, and so on. At the end of the inference process for a given series of encoder barcode sequences, the binder identifier string is actually generated because all binder identifiers are detected and analyzed in a specific order during the inference process, an order determined by the order of specific coding events that have occurred for a specific polypeptide in the coding assay (the order of binders encoding a specific polypeptide in the coding assay).

[0284] In some embodiments of the disclosed method, the computer model is trained based on predetermined parameters of the coding assay, wherein the predetermined parameters include: i) the probability of correctly or incorrectly assigning one or more binder identifiers to at least one component of a given polypeptide or polypeptide fragment based on determined binding characteristics of the binders used in the coding assay; and ii) optionally, the probability of successfully cleaving at least one component of a given polypeptide or polypeptide fragment.

[0285] In some embodiments, to infer the amino acid sequence of a polypeptide from a conjugate identifier, the binding characteristics of the conjugate are determined experimentally by testing the binding of various conjugates to a variety of known peptides (e.g., testing the binding affinity of the conjugate using an array of different peptides). In other embodiments, the binding characteristics of the conjugate are predicted based on computer modeling of the conjugate's structure and analysis in a computer of the interactions between the conjugate and a variety of known peptides. In yet another embodiment, a combination of methods (both computational and experimental) can be used to determine the binding characteristics of the conjugate.

[0286] In some embodiments, the decoding algorithm infers the amino acid sequence of a polypeptide from multiple polypeptides from the binder identifiers in the binder identifier strings, the inference being based on (i) the binding characteristics of binders from a set of binders corresponding to binder identifiers in the binder identifier strings, and (ii) a probability score calculated for the association between one or more binder identifiers in the binder identifier strings and one or more amino acid sequences of the polypeptides in the multiple polypeptides.

[0287] In some embodiments, the decoding algorithm calculates a probability score of the association between one or more binder identifiers in a binder identifier string and one or more amino acid sequences of a polypeptide. This calculation uses (i) binding features from a set of binders corresponding to one or more binder identifiers; and (ii) one or more parameters of an encoding assay. In some embodiments, the one or more parameters of the encoding assay include the efficiency of functionalization of N-terminal amino acid (NTAA) residues of polypeptides in multiple polypeptides, the efficiency of cleavage of NTAA residues of polypeptides in multiple polypeptides, the efficiency of encoding of NTAA residues of polypeptides in multiple polypeptides, or any combination thereof.

[0288] In some embodiments, the term "association" between one or more binding identifiers and one or more amino acid sequences of a polypeptide refers to the fact that one or more amino acid sequences of a polypeptide generate one or more binding identifiers during coding assays and subsequent post-coding assays of multiple nucleic acid sequences (outputs of coding assays). A calculated probability score illustrates the likelihood of a "correct" association. For example, when the probability score is 100%, there is 100% confidence that the computer model correctly associates one or more amino acid sequences of a polypeptide from multiple polypeptides with one or more binding identifiers from multiple binding identifier strings. If the probability score is 70%, there is a 70% probability that the association is correct and a 30% probability that the association is incorrect. In some embodiments, even when the probability score is less than 100%, the computer model can still infer one or more amino acid sequences of a polypeptide from a given binding identifier string. In some embodiments, the computer model outputs certain amino acid sequences of polypeptides from multiple polypeptides associated with one or more binding identifiers from multiple binding identifier strings, and further provides a probability score for the presence of each amino acid sequence within multiple polypeptides.

[0289] In some embodiments, the method disclosed above further includes the following steps performed after (a) and (b): (d) for each of a plurality of polypeptides and / or polypeptide fragments, generating a set of binder identifier strings based on predetermined parameters of a coding assay, wherein the predetermined parameters include: i) the probability of correctly or incorrectly assigning one or more binder identifiers to at least one component of a given polypeptide or polypeptide fragment based on the binding characteristics of the binders used in the coding assay; and ii) optionally, the probability of successfully cleaving at least one component of a given polypeptide or polypeptide fragment, thereby determining the probability of each binder identifier string generated from the given polypeptide or polypeptide fragment; and (e) for each of the plurality of binder identifier strings generated in (b), calculating the probability of a given binder identifier string being generated from one or more polypeptides or fragments thereof based on the probability of the binder identifier strings determined in (d), thereby determining at least partial identity and / or quantity of at least one polypeptide among a plurality of polypeptides.

[0290] In some embodiments, a set of binder identifier strings based on predetermined parameters of a coding assay is generated not only for each of a variety of polypeptides and / or polypeptide fragments, but also for at least one other polypeptide and / or polypeptide fragment that may be present in the analyzed sample. For example, when the sample is human plasma, all polypeptides that may be present in human plasma can be processed to generate a set of binder identifier strings and determine the probability of each binder identifier string being generated from a given polypeptide or polypeptide fragment. Each individual plasma sample may contain only a small subset of plasma proteins (other proteins may be, for example, degraded), and it is unknown which proteins are present in the sample and which are not. Therefore, it is necessary to generate a set of binder identifier strings for all polypeptides that may be present in the sample, thereby creating a large table with (simulated) binder identifier strings and the probability of each (simulated) binder identifier string being generated from a given polypeptide or polypeptide fragment. The data generated from the coding assay is then matched against this table, and the polypeptides of a particular sample are identified or at least predicted with a certain probability.

[0291] In some embodiments, a set of combiner identifier strings based on predetermined parameters of an encoding determination method is generated in a computer, such that it is a set of simulated combiner identifier strings, and the probability of each simulated combiner identifier string is determined. Using simulated combiner identifier strings allows for the generation of significantly more data than experimentally generated data, which is extremely useful for training purposes.

[0292] In some embodiments, a set of combiner identifier strings is a set of simulated combiner identifier strings generated in a computer in (d), wherein the probability of each simulated combiner identifier string is determined in (d). In other embodiments, a set of combiner identifier strings is generated experimentally using a coding determination method. In yet another embodiment, a combination method (both computationally and experimentally) can be used to generate a set of combiner identifier strings based on predetermined parameters of a coding determination method.

[0293] In some embodiments of the disclosed method, the multiple polypeptides encoded by a plurality of nucleic acid sequences by the coding assay include at least 1,000, at least 10,000, at least 100,000, at least 1,000,000 or more polypeptides.

[0294] In some embodiments of the disclosed method, the multiple polypeptides encoded by a plurality of nucleic acid sequences via the coding assay include at least 500, at least 1000, at least 2000, at least 5000, at least 10,000, at least 20,000 or more different polypeptides. In some embodiments, the different polypeptides are derived from 2, 5, 10, 50, 100, 200 or more different biological samples analyzed in a single coding assay.

[0295] In some embodiments of the disclosed method, the series of encoder barcode sequences includes at least two different encoder barcode sequences. In some embodiments of the disclosed method, the series of encoder barcode sequences includes no more than 30 different encoder barcode sequences. In some embodiments of the disclosed method, the series of encoder barcode sequences includes 4 to 20 different encoder barcode sequences (each unique barcode sequence in the series of encoder barcode sequences corresponds to a unique binder of a component of a polypeptide that binds to multiple polypeptides in the coding assay). In some embodiments, no more than 30 unique binders are required to fully characterize at least 1,000, at least 10,000, at least 100,000, at least 1,000,000 or more polypeptides.

[0296] In some embodiments of the disclosed method, the components of the polypeptide bound by the binder consist of terminal amino acid residues modified with an N-terminal modifier.

[0297] In some embodiments of the disclosed method, predetermined parameters are obtained by training a machine learning model that uses output data generated from multiple coding assays with predetermined analytes as input to generate a computer model.

[0298] In some embodiments of the disclosed method, the computer model outputs the probability that each binder identifier string originates from two or more amino acid sequences of a polypeptide associated with a given binder identifier string.

[0299] In some embodiments of the disclosed method, i) each of the plurality of nucleic acid sequences further includes one or more auxiliary sequences; ii) the method further includes using one or more processors to assign auxiliary identifiers to at least one of the one or more auxiliary sequences in each of the plurality of nucleic acid sequences to generate a plurality of identifier strings for the plurality of nucleic acid sequences; and iii) the one or more auxiliary sequences include one or more identifier sequences or their complements, one or more spacer sequences or their complements, one or more sequencing primer sequences or their complements, or any combination thereof.

[0300] In some embodiments of the disclosed method, one or more identifier sequences include one or more sample barcode sequences, one or more bead barcode sequences, one or more unique molecular identifier (UMI) sequences, or any combination thereof.

[0301] In some embodiments of the disclosed method, the assignment of binder identifiers to the encoder barcode sequence is based on a specific order of components in each of a plurality of nucleic acid sequences and a lookup table (LUT) for binder identifiers.

[0302] In some embodiments of the disclosed method, the assignment of sub-identifiers to the encoder barcode sequence further includes error correction based on an enumeration of error types to account for errors associated with the encoding process used to generate each encoder barcode sequence in a series of encoder barcode sequences.

[0303] In some embodiments of the disclosed method, considering errors associated with the process of generating each encoder barcode sequence in a series of encoder barcode sequences includes determining: (i) a first probability of reading a specified barcode sequence of a specified underlying polypeptide component, and (ii) a second probability of successfully transferring from one polypeptide component to the next during the encoding process.

[0304] In some embodiments of the disclosed method, the assignment of the binder identifier to the encoder barcode sequence in (b) includes using a probabilistic model to predict the sequence of binder identifiers for each of a plurality of nucleic acid sequences.

[0305] In some embodiments of the disclosed method, the probabilistic model includes a Hidden Markov Model (HMM). In some embodiments, the HMM is trained using one or more training datasets comprising tag pairs of binder identifiers and nucleic acid sequences. In some embodiments, the HMM is trained using an iterative expectation-maximization (EM) algorithm to determine a set of model parameters that maximize the probability of correctly predicting the sequence of binder identifiers for each of a plurality of nucleic acid sequences.

[0306] In some embodiments of the disclosed method, the expectation-maximization (EM) algorithm includes the Baum-Welch algorithm.

[0307] In some embodiments of the disclosed method, a computer model is configured to identify a unique binder identifier feature spectrum among a plurality of binder identifier strings as part of inferring the amino acid sequence of a polypeptide from the binder identifier strings, and wherein a given unique binder identifier feature spectrum comprises a set of binder identifier strings associated with a single polypeptide.

[0308] In some embodiments of the disclosed method, the computer model is trained using a training dataset that includes binder identifier string data of one or more isolated peptide samples processed using the same sample preparation protocol used for processing multiple peptides.

[0309] In some embodiments of the disclosed method, the training dataset further includes binder identifier string data of background samples processed using the same sample preparation protocol as that used for processing multiple peptides.

[0310] In some embodiments of the disclosed method, the background sample includes a plasma sample, a urine sample, a saliva sample, or a cell extract sample.

[0311] In some embodiments of the disclosed method, the sample is a cell extract sample, and the cell extract sample includes mammalian cell extract samples, plant cell extracts, fungal cell extracts, or bacterial cell extract samples.

[0312] In some embodiments of the disclosed method, the computer model is further configured to correct the quantity output of at least one polypeptide using a correction factor computed from the training dataset.

[0313] In some embodiments of the disclosed method, the computer model is configured to map each of a plurality of binder identifier strings to a specific polypeptide as part of assigning the binder identifier string to an amino acid sequence.

[0314] In some embodiments of the disclosed method, mapping a binder identifier string from a plurality of binder identifier strings to a specific polypeptide includes: i) generating a set of k-mer fragments of binder identifier strings; ii) determining the probability that a given k-mer fragment in the set belongs to a specific polypeptide based on a previously determined probability distribution; and iii) assigning the binder identifier string to the specific polypeptide based on the determined probability of the set of k-mer fragments.

[0315] In some embodiments of the disclosed method, the probabilistic model is trained on measurement parameter data determined empirically.

[0316] In some embodiments of the disclosed method, for each barcode sequence in a series of encoder barcode sequences, empirically determined assay parameter data includes the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of the corresponding polypeptide bound by the binder.

[0317] In some embodiments of the disclosed method, the assay parameter data determined empirically for each potential N-terminal amino acid residue in the polypeptide includes the probability of successfully cleaving the N-terminal amino acid residue during a cycling process used to encode multiple polypeptides in multiple nucleic acid sequences.

[0318] In some embodiments of the disclosed method, mapping a binder identifier string from a plurality of binder identifier strings to a specific polypeptide includes: providing a plurality of binder identifier strings as input to a computer model, wherein the computer model is configured to assign a given binder identifier string to two or more specific polypeptides in a fractional manner.

[0319] In some embodiments of the disclosed method, the computer model is trained on a training dataset comprising a set of simulated binder identifier strings generated based on empirically determined assay parameter data for a given input distribution of a peptide, the assay parameter data including: i) the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of the corresponding peptide bound by the binder for each barcode sequence in a series of encoder barcode sequences; and ii) the probability of successfully cleaving an N-terminal amino acid residue for each potential N-terminal amino acid residue in the peptide during a cyclic process for encoding multiple peptides in multiple nucleic acid sequences.

[0320] In some embodiments of the disclosed method, a set of simulated binder identifier strings includes at least 100,000, 500,000, 1M, 5M, 10M, 25M, 50M, 75M, 100M, 200M, 300M, 400M, or 500M simulated binder identifier strings.

[0321] In some embodiments of the disclosed method, the computer model is further configured to output confidence intervals for partial identity of at least one polypeptide.

[0322] In some embodiments of the disclosed method, the stringency level of the output confidence interval can be selected by the user.

[0323] In some embodiments of the disclosed method, the stringency level of the confidence interval corresponds to a confidence level of 90%, 95%, 98%, or 99%.

[0324] In some embodiments of the disclosed method, the computer model includes a trained artificial neural network model, a trained deep learning model, a trained random forest model, or a trained support vector machine.

[0325] In some embodiments, the disclosed method further includes performing an iterative expectation-maximization (EM) process to refine the quantity output of at least one polypeptide among a plurality of polypeptides.

[0326] In some embodiments of the disclosed method, the iterative EM process repeatedly includes: (i) finding the optimal allocation of the binder to each barcode sequence in the nucleic acid sequence based on a current estimate of which peptides are present among a variety of peptides; (ii) finding an updated optimal estimate of which peptides are present among the variety of peptides based on the optimal allocation of the binder to each different barcode sequence in the nucleic acid sequence; and (iii) determining the amount of at least one peptide present among the variety of peptides based on the updated optimal estimate of which peptides are present among the variety of peptides. In some embodiments, steps (i) to (iii) are repeated until the difference between the amount of at least one peptide determined in one iteration and the next is less than a specified threshold, or until a specified maximum number of iterations has been reached.

[0327] In some embodiments, the disclosed method further includes determining the nucleic acid sequences of a plurality of nucleic acid sequences. In some embodiments, the nucleic acid sequences are determined by performing DNA sequencing. In some embodiments, the DNA sequencing is performed using a next-generation DNA sequencer.

[0328] In some embodiments of the disclosed method, the plurality of nucleic acid sequences includes at least 10,000, 100,000, or 1M nucleic acid sequences.

[0329] In some embodiments of the disclosed method, the multiple polypeptides include at least 100, 500, 1,000, 2,000, 5,000, or 10,000 different polypeptides.

[0330] In some embodiments of the disclosed method, at least two binder identifier strings are generated for at least one of a plurality of nucleic acid sequences.

[0331] In some embodiments, the disclosed method further includes comparing a plurality of nucleic acid sequences output by a trained model with a plurality of nucleic acid sequences determined by subjecting a plurality of peptides to a reverse translation assay and sequencing the resulting extended recording tags.

[0332] In some embodiments of the disclosed method, the computer model includes a machine statistical model. In other embodiments, the computer model is a trained machine learning model.

[0333] In a preferred embodiment, the disclosed method is used for the simultaneous analysis and / or sequencing (multiplexing) of multiple peptide analytes.

[0334] As used in this article, multiplexing refers to the analysis of multiple peptide analytes in the same assay. Multiple peptide analytes can originate from the same or different samples. Multiple peptide analytes can originate from the same or different subjects. The analyzed multiple peptide analytes can be different peptide analytes or the same peptide analyte originating from different samples. Multiple peptide analytes include 2 or more peptide analytes, 5 or more peptide analytes, 10 or more peptide analytes, 50 or more peptide analytes, 100 or more peptide analytes, 500 or more peptide analytes, 1000 or more peptide analytes, 5,000 or more peptide analytes, 10,000 or more peptide analytes, 500,000 or more peptide analytes, or 1,000,000 or more peptide analytes.

[0335] In a preferred embodiment, the disclosed method is used to identify a large number of peptides (e.g., at least 1,000, 10,000, 100,000, 1,000,000 or more peptide molecules, including at least 100, 1,000, 10,000 or more different peptide molecules) in a single assay.

[0336] In some embodiments of the disclosed method, the molecules include two or more molecules of the same polypeptide analyte and two or more molecules of different polypeptide analytes.

[0337] In some embodiments of the disclosed methods and compositions, any two or more molecules are independently associated with the same nucleic acid recording tag or different nucleic acid recording tags.

[0338] In some embodiments of the disclosed methods and compositions, any two or more of the different peptide analytes are independently associated with the same nucleic acid recording tag or different nucleic acid recording tags.

[0339] In some embodiments, this document provides a method for identifying a large number of peptide analytes (e.g., at least 1,000, 10,000, 100,000, 1,000,000 or more peptide molecules, including at least 100, 1,000, 10,000 or more different peptide molecules) in a single assay. This method uses one or more deconvolution methods to match a set of binders with the sequence of the peptide in each of the multiple peptide analytes based on the determined binding properties of the binders, thereby determining the identity of each peptide analyte among the multiple peptide analytes. In a preferred embodiment, the known specificity (binding characteristics) of the binders to NTAA residues and the order in which they bind to the peptide analytes are used to decode the identity of the peptide analytes.

[0340] In preferred embodiments, the methods provided herein are capable of simultaneously identifying multiple different peptide analytes (such as at least 100, 1000, 10000 or more different peptides) within a single sample. In some embodiments, peptides from a sample may be fractionated into multiple fractions, and peptides in each fraction may be fragmented into peptides, which are then barcoded (e.g., by incorporating a sample barcode into an associated record tag for each peptide). The barcoded peptides from different fractions, each concatenated to a record tag, can then be merged together and analyzed using the methods and compositions disclosed herein. Fractionation, barcoding, and merging techniques are advantageous for analyzing complex biological samples, such as samples with peptides exhibiting significant abundance differences (e.g., plasma). Techniques for fractionation, barcoding, and merging are known in the art and disclosed, for example, in US 20190145982 A1, which is incorporated herein by reference.

[0341] In some embodiments of the disclosed method, given that the selectivity of each binder in the binders to NTAA residues is known, information regarding the identity of NTAA residues of the immobilized peptide being analyzed is encoded in a unique nucleic acid barcode present in the extended recording tag. This nucleic acid barcode can be used to decode the identity of NTAA residues using known information about the binding kinetics and / or specificity of the binders that bind to the peptide at a given binding cycle. In some embodiments, the nucleic acid barcode can be used as input to a probabilistic neural network trained to associate the sequence of the barcode with amino acid identity. Training can be performed by independently testing each binder (optionally concatenated to the coding tag) against a set of peptides, each with different NTAA residues and associated recording tags, collecting sequence information of the extended recording tag after binding, and feeding the collected information into the probabilistic neural network. Alternatively, training can be performed by testing a set of binders (optionally each concatenated to the coding tag) against a set of peptides, collecting sequence information of the extended recording tag after binding, and feeding the collected information into the probabilistic neural network.

[0342] In some embodiments, during each encoding cycle, only a single amino acid residue of the analyzed peptide is encoded into the recording tag (each time it is an NTAA residue, which is cleaved at the end of each binding cycle). In other embodiments, a dipeptide is encoded into the recording tag and the dipeptide is cleaved between encoding cycles (e.g., by using a dipeptidyl carboxypeptidase).

[0343] In some embodiments, the DNA barcode generated on the extended recording tag of each peptide analyte is input into a probabilistic neural network (PNN), which learns to associate the sequence of the DNA barcode with amino acid identity. Probabilistic neural networks (Mohebali, B., et al., Chapter 14 – Probabilistic neural networks: a brief overview of theory, implementation, and application, in Handbook of Probabilistic Models, P. Samui, et al., ed. 2020, Butterworth-Heinemann. pp. 347-367) approach Bayesian optimal classification for multi-class problems such as amino acid identification from DNA barcodes (Klocker, J., et al., Bayesian Neural Networks for Aroma Classification. Journal of Chemical Information and Computer Sciences, 2002. 42(6): pp. 1443-1449). As the size of the representative dataset increases, PNN-based classifiers are guaranteed to learn and converge to an optimal classifier. Probabilistic neural networks have a parallel structure, which allows data from any amino acid residue to be used to learn / predict most other amino acid residues.

[0344] In some embodiments, the disclosed method is used for determining peptide sequences based on a probabilistic neural network ensemble. This machine learning method is characterized by sequence determination achieved through the following steps: i) encoding peptide fragments of a peptide into fragments of a DNA sequence using a set of binders based on the structural properties of amino acid residues; ii) establishing a set of probabilistic neural network subclassifiers, using peptide fragments of peptides with known sequences for amino acid classification training, and obtaining a set of trained amino acid classification models; iii) using the obtained models to determine the amino acid sequences of peptides in a test dataset; iv) counting the classification results output by the models to generate an amino acid candidate set; v) combining methods demonstrating the highest accuracy to determine the amino acid sequences of peptide fragments; and vi) validating the amino acid determination results of the algorithm using k-fold cross-validation, where k is an integer.

[0345] In some embodiments, k-fold cross-validation operates as follows. In k-fold cross-validation, the dataset is shuffled and randomly divided into k groups with no overlap and replacement. This means that each group is unique and used only once for model evaluation. The data groups complete the following steps for k-fold cross-validation:

[0346] 1) Select a unique group as the test dataset;

[0347] 2) The remaining (k-1) groups are used as the training dataset;

[0348] 3) The model is built using the training set and evaluated using the test set;

[0349] 4) Retain the evaluation score and discard the model;

[0350] 5) Repeat steps 1 to 4 until all k groups have been used for model evaluation;

[0351] 6) Output the average of the k evaluation scores as the model performance score for k-fold cross-validation.

[0352] In some embodiments, the nucleic acid barcode on the extended recording tag of each peptide analyte is fed into a probabilistic neural network (PNN), which learns to associate the DNA sequence of the barcode with the amino acid identity of the peptide being analyzed. In other embodiments, other statistical models (e.g., hidden Markov models) and machine learning methods (e.g., random forest models) can be used to classify NGS reads from each extended recording tag into specific amino acid residues (or amino acid residue types, if the binder is not selective for a specific NTAA residue).

[0353] In some embodiments of the disclosed method, the N-terminal amino acid residues of the peptide analyte are not modified or removed.

[0354] In some embodiments of the disclosed method, the N-terminal amino acid (NTAA) residues of the peptide analyte are modified with an N-terminal modifier. In some embodiments of the disclosed method, the N-terminal modifier is a compound selected from the group consisting of:

[0355]

[0356] Where M is a metal-bonded group containing sulfonamide, isohydroxamic acid, aminosulfonate, or thioamide; group It is a 5- or 6-membered aromatic ring containing up to three heteroatoms selected from N, O, and S as ring members, and optionally substituted with R; R represents one or two optional substituents selected from the group consisting of: F, Cl, CH3, CF2H, CF3, OH, OCH3, OCF3, NH2, N(CH3)2, NO2, SCH3, SO2CH3, CH2OH, B(OH)2, CN, CONH2, CO2H, CN4H, and CONHCH3; LG is OH, ORQ, or OCC, each RQ is independently aryl or heteroaryl, each optionally substituted with one or more groups selected from halogen, nitro, cyano, sulfonate, carboxylate, and alkylsulfonyl, and the N of the heteroaryl group is optionally oxidized; or RQ can be -C(=O)R or -C(=O)-OR; CC is a cationic counterion; X is one of: O, S, Se, or NH. In some embodiments of the disclosed method, the N-terminal modifier is a compound of formula (5):

[0357] (5)

[0358] in:

[0359] G 1 To G 4 Each is independently selected from CH, CJ, and N, provided that G is selected. 1 To G 4 No more than 3 of them are N;

[0360] J is independently selected from H, C1-C2 alkyl, NO2, C1-C2 haloalkyl, C1-C2 haloalkoxy, halogen, and -OR each time it appears. 8 -N(R) 8 )2、-SR 8 -S(O) n R 8 -NR 8 SO2R 8 -SO2N(R) 8 2. SO3R 8 -B(OR) 8 2. C(=O)R 8 CN, CON(R) 8 )2、-COOR 8 -C(-O)Ar and tetrazolium, wherein Ar represents a phenyl group or is optionally selected from halogens, CN, R 8 and OR 8 A 5- to 6-membered heteroaryl ring substituted with one or two groups;

[0361] Each R 8 Independently selected from H, C1-C2 alkyl, and C1-C2 haloalkyl;

[0362] n is independently 1 or 2 each time it appears; and

[0363] R 13 The N-terminal modifier is selected from H, C1-C2 alkyl, C1-C2 alkoxy, C1-C2 haloalkyl, and C1-C2 haloalkoxy. In some embodiments of the disclosed method, the N-terminal modifier is a compound described in the following patent disclosures: US 2023 / 0220589 A1 and US 2022 / 0283175 A1, which are incorporated herein by reference.

[0364] Encoded tag information associated with a specific binder can be transferred to a recording tag using a variety of methods. In any of the disclosed embodiments, the transfer of information (e.g., from the encoding tag to the recording tag, or from the recording tag to the encoding tag) can be accomplished by ligation (e.g., enzymatic or chemical ligation, splint ligation, sticky end ligation, single-strand (ss) ligation (such as ssDNA ligation) or any combination thereof), polymerase-mediated reactions (e.g., primer extension of single-stranded or double-stranded nucleic acids) or any combination thereof.

[0365] In some embodiments, the DNA polymerase used for primer extension during information transfer has strand displacement activity and limited or absent 3'-5 exonuclease activity. Several examples of such polymerases include Klenowexo- (Klenow fragment of DNA Pol 1), T4 DNA polymerase exo-, T7 DNA polymerase exo- (sequencing enzyme 2.0), Pfuexo-, Vent exo-, Deep Vent exo-, Bst DNA polymerase large fragment exo-, Bca Pol, and Phi29 Polexo-. In a preferred embodiment, the DNA polymerase is active at room temperature and up to 45°C. In another embodiment, a "warm-start" version of a thermophilic polymerase is employed, such that the polymerase is activated and used at approximately 40°C to 50°C. An exemplary warm-start polymerase is Bst 2.0 warm-start DNA polymerase (New England Biolabs).

[0366] Encoding tag information associated with a specific conjugate can be transferred via ligation to a recording tag associated with the immobilized polypeptide. Ligation can be blunt-end ligation or sticky-end ligation. Ligation can be an enzymatic ligation reaction. Examples of ligases include, but are not limited to, CV DNA ligase, T4 DNA ligase, T7 DNA ligase, T3 DNA ligase, Taq DNA ligase, and E. coli DNA ligase. Alternatively, ligation can be a chemical ligation reaction. In one embodiment, spacerless ligation is accomplished by hybridization of a “recording helper” sequence with an arm on the encoding tag. Annealed complementary sequences are chemically ligated using standard chemical ligation or “click chemistry” (Gunderson et al., Genome Res (1998) 8(11): 1142-1153; Litovchick et al., Artif DNA PNA XNA (2014) 5(1): e27896; Roloff et al., MethodsMol Biol (2014) 1050:131-141).

[0367] Various aspects of the coded label and record label composition, as well as aspects of transferring identification information from the coded label to the record label, are disclosed in public applications US 2019 / 0145982 A1 and US 2023 / 0136966 A1, which are incorporated herein by reference.

[0368] In some embodiments, the coded tags within a set of binding reagents share a common spacer sequence used in the assay (e.g., the entire library of binding reagents used in a multi-binding-cycle method has a common spacer in its coded tag). In another embodiment, the coded tag includes a binding-cycle-specific barcode that identifies a particular binding cycle. In other embodiments, the coded tags within a set of binding reagents have a binding-cycle-specific spacer sequence. For example, the coded tag for the binding reagent used in a first binding cycle includes a spacer sequence specific to "Cycle 1", the coded tag for the binding reagent used in a second binding cycle includes a spacer sequence specific to "Cycle 2", and so on, up to "n" binding cycles. In a further embodiment, the coded tag for the binding reagent used in a first binding cycle includes a spacer sequence specific to "Cycle 1" and a spacer sequence specific to "Cycle 2", the coded tag for the binding reagent used in a second binding cycle includes a spacer sequence specific to "Cycle 2" and a spacer sequence specific to "Cycle 3", and so on, up to "n" binding cycles. This embodiment is useful for subsequent PCR assembly of non-tandem extended recording tags after the binding cycle is completed. In some embodiments, the spacer sequence comprises a sufficient number of bases to anneal with a complementary spacer sequence in the recording tag or extended recording tag to initiate a primer extension reaction or a sticky end joining reaction.

[0369] In some embodiments, each peptide may employ a set of recording tags to improve the overall robustness and efficiency of encoding tag information transfer. Using a set of recording tags associated with a given peptide, rather than a single recording tag, improves the efficiency of library construction because the coupling yield between encoding tags and recording tags can be higher, and the overall library yield is higher. The yield of a single tandemly extended recording tag depends directly on the stepwise yield of tandem, while using multiple recording tags capable of accepting encoding tag information does not suffer from the exponential loss of tandem.

[0370] In some embodiments of the disclosed method, the N-terminal amino acid residues of each polypeptide analyte are linked to a support or a nucleic acid recording tag associated with the polypeptide analyte.

[0371] In some embodiments of the disclosed method, the analytical steps include determining the identity of the NTAA residue and the penultimate terminal amino acid of the peptide analyte.

[0372] The conjugates in the disclosed methods and compositions do not need to be strictly selective and capable of recognizing, for example, functional classes of NTAA residues, such as negatively charged residues, positively charged residues, small hydrophobic residues, aromatic residues, etc., or recognizing other NTAA residue types. In some embodiments, at least some of the conjugates in a group are degenerate (each capable of binding more than one structure or more components of the polypeptide). In some embodiments, degenerate conjugates are specific to two or more NTAA residues. In some embodiments, each conjugate in the group is not highly specific to a particular NTAA. Using degenerate conjugates reduces the total number of conjugates required for successful polypeptide identification. In some embodiments, identifying at least 90% of the polypeptide analytes present in a sample requires no more than 5, 6, 7, 8, 9, or 10 conjugates with different NTAA specificities.

[0373] In some embodiments, the peptide analyte may be mixed, spotted, dropped, pipetted, flowed, or otherwise applied to a support. In some embodiments, the support has been chemically functionalized (e.g., an NHS ester or other amine-specific reagent) before the peptide analyte is applied to it. This enables the peptide analyte to be immobilized to the support via its N-terminus (see also Example 3 below).

[0374] In a preferred embodiment, the selectivity of each binder for the NTAA residues of the peptide analyte is determined in advance prior to the contact steps of the disclosed method. Each binder may be tested against a group of peptides with different NTAA residues and associated recording tags to characterize the selectivity of the binder for each of the 20 natural NTAA residues and optionally the binding kinetics. When multiple alternative binders are available, a minimum number of binders may be selected that will cover all or the maximum number of the 20 natural NTAA residues.

[0375] Attached to support

[0376] In some embodiments, the peptide is attached to a support prior to the binding / cleavage reaction. In some cases, it is necessary to immobilize a large number of peptides using a support with high load-bearing capacity. In some embodiments, it is preferable to use a three-dimensional support (e.g., a porous matrix or porous beads) to immobilize the peptide. For example, the preparation of the peptide, including the attachment of the peptide to the support, is performed before the binding reaction. In some instances, the preparation of the peptide, including the attachment of the peptide to a nucleic acid molecule or oligonucleotide, may be performed before or after the immobilization of the peptide. In some embodiments, multiple peptides are attached to the support prior to the binding reaction and contact with the conjugate.

[0377] In some embodiments, the support may comprise any suitable solid material, including porous and nonporous materials, and the polypeptide (e.g., a peptide) may be directly or indirectly associated with the material by any means known in the art, including covalent and nonvalent interactions or any combination thereof.

[0378] Various reactions can be used to attach peptide analytes to a support, or to associate each peptide analyte with a recording tag (directly or indirectly, such as via the support). Peptides can be attached to the support directly or indirectly. In some cases, peptides are attached to the support via nucleic acids (e.g., via nucleic acid recording tags). Exemplary reactions include click chemistry reactions such as copper-catalyzed reactions of azides and alkynes to form triazoles (Huisgen 1,3-dipolar cycloaddition), strain-promoted azide-alkyne cycloaddition (SPAAC), reactions of dienes and dienophiles (Diels-Alder), strain-promoted alkyne-nitroketone cycloaddition, reactions of strained alkenes with azides, tetrazides, or tetrazolium, cycloaddition of alkenes with azides [3+2], electron-demanding Diels-Alder (IEDDA) reactions of alkenes and tetrazides (e.g., m-tetrazide (mTet) or phenyltetrazide (pTet) and trans-cyclooctene (TCO); or pTet and alkenes), photoreactions of alkenes and tetrazolium, Staudinger linkages of azides and phosphine, and various substitution reactions such as substitution of leaving groups by nucleophilic attack on electrophilic atoms (Horisawa 2014, Knall, Hollauf et al. 2014). Exemplary substitution reactions include reactions of amines with: activated esters; N-hydroxysuccinimide esters; isocyanates; isothiocyanates, aldehydes, epoxides, etc. In some embodiments, iEDDA click chemistry is used to immobilize peptides to a support because it is rapid and provides high yields at low input concentrations. In another embodiment, m-tetrazine is used in the iEDDA click chemistry reaction instead of tetrazine because m-tetrazine has improved bond stability. In another embodiment, phenyltetrazine (pTet) is used in the iEDDA click chemistry reaction. In one case, the peptide is labeled with a bifunctional click chemistry reagent, such as an acetylene-NHS ester (acetylene-PEG-NHS ester) reagent or acetylene-benzophenone, to generate an acetylene-labeled peptide. In some embodiments, the alkyne may also be a strained alkyne, such as cyclooctyne, including dibenzocyclooctyl (DBCO). Other suitable examples of covalent conjugation between two moieties are disclosed in US 2021 / 0101930 A1, which is incorporated herein by reference.

[0379] Similar methods (e.g., click chemistry, bioorthogonal reactions) can be used to attach peptide analytes to associated nucleic acid recording tags or to binders to associated nucleic acid coding tags. Such attachments can be achieved by introducing reactive moieties on one or both attachment partners.

[0380] In some embodiments of the disclosed method, a variety of different peptides are immobilized on a solid support, each of the various different peptides being associated with a nucleic acid recording tag. There are several possible ways to establish the association between the immobilized peptide and the associated nucleic acid recording tag. The recording tag may be directly attached to the peptide, via a linker, via a multifunctional linker, or associated with the peptide by means of its proximity (or co-localization) on the support. In some embodiments, the recording tag is attached to the support, and the peptide is immobilized on the support via the recording tag. In some embodiments, a linker is attached to the support, and the peptide and the recording tag are independently attached to the linker, thereby generating immobilization on the support and association between the peptide and the recording tag. Other variations of immobilization and association are also possible.

[0381] In some embodiments, at least one recording tag is directly or indirectly associated with or co-located to the peptide. In another embodiment, multiple recording tags are attached to the peptide, such as to lysine residues or the peptide backbone. In some embodiments, the peptide labeled with multiple recording tags is fragmented or digested into smaller peptides, with an average of one recording tag per peptide. Recording tags may be single-chain or partially or fully double-chain. In some embodiments, recording tags may include unique molecular identifiers, compartment tags, partition barcodes, sample barcodes, fractionation barcodes, spacer sequences, universal initiation sites, or any combination thereof. In some embodiments, recording tags may include blocking groups, such as at the 3' end of the recording tag. In some cases, the 3' end of the recording tag is blocked to prevent the recording tag from being extended by polymerase.

[0382] In some embodiments, the record tag may include a sample identification barcode. Sample barcodes are useful in multiplexing a set of samples in a single reaction vessel or immobilized to a single solid support (e.g., beads or planar matrix) or a collection of solid supports. For example, peptides from many different samples can be tagged with record tags having sample-specific barcodes, and then all samples are merged together before immobilization to the support, binding of binders, and record tag analysis. In some embodiments, the record tag includes an optional unique molecular identifier (UMI), which provides a unique identifier tag for each peptide associated with the UMI. The UMI can be used to deconvolve sequencing data from multiple extended record tags to identify sequence reads from individual peptides. In some embodiments, within a peptide library, each peptide is associated with a single record tag, and each record tag includes a unique UMI. In other embodiments, multiple copies of the record tag are associated with a single peptide, and each copy of the record tag includes the same UMI. In some embodiments, the record tag includes a universal priming site, such as a forward or 5' universal priming site. A universal priming site is a nucleic acid sequence that can be used to initiate a library amplification reaction and / or for sequencing. Universal initiation sites may include, but are not limited to, PCR amplification initiation sites, flow cell adaptor sequences (which anneal with complementary oligonucleotides on the flow cell surface (e.g., Illumina next-generation sequencing)), sequencing initiation sites, or combinations thereof. In some embodiments, universal initiation sites include Illumina P5 primers or Illumina P7 primers for NGS.

[0383] In some embodiments, the identification of components of an encoded tag or record tag (e.g., barcode, UMI, compartment label, partition barcode, sample barcode, spatial region barcode, cycle-specific sequence, or any combination thereof) undergoes Hamming distance, Lie distance, asymmetric Lie distance, Reed-Solomon, Levenshtein-Tenengolts, or similar error correction methods. Hamming distance refers to the number of distinct positions between two strings of equal length. It measures the minimum number of substitutions required to change one string to the other. Hamming distance can be used for error correction by selecting encoder sequences that are reasonably spaced. Thus, in an example where the encoder sequence is 5 bases, the number of available encoder sequences is reduced to 256 unique encoder sequences (Hamming distance 1→4). 4One encoder sequence = 256 encoder sequences. In another embodiment, the encoder sequence, barcode, UMI, compartment label, cycle-specific sequence, or any combination thereof is designed to be easily read out by a cycle decoding process (Gunderson, 2004, Genome Res. 14:870-7). In another embodiment, the barcode, UMI, compartment label, partition barcode, spatial barcode, sample barcode, cycle-specific sequence, or any combination thereof is designed to be read out by low-accuracy nanopore sequencing because instead of single-base resolution, it is necessary to read words of multiple bases (approximately 5 to 20 bases in length).

[0384] In some embodiments where multiple peptides are immobilized on the same support, the peptide molecules may be appropriately spaced to accommodate methods for performing binding reactions and any downstream analytical steps for evaluating the peptides. For example, it may be advantageous to optimally space the peptide molecules to enable nucleic acid-based methods for evaluating and sequencing the peptides. To control the spacing of the immobilized peptides on the support, the density of functional coupling groups (e.g., TCO or carboxyl groups (COOH)) used to attach the peptides may be titrated on the support surface. In some embodiments, multiple peptide molecules are spaced apart on the surface or within the volume of the support (e.g., for porous supports) such that adjacent molecules are spaced apart at a distance of about 50 nm to about 500 nm. Further details of these methods can be found in published applications US 2019 / 0145982 A1 and US 2023 / 0136966 A1, which are incorporated herein by reference.

[0385] Exemplary embodiments

[0386] The following exemplary embodiments represent certain embodiments and examples of the present invention:

[0387] 1. A computer-implemented method for analyzing multiple polypeptides encoded in multiple nucleic acid sequences by a coding assay, the method comprising:

[0388] (a) Receives at one or more processors a plurality of nucleic acid sequences generated from an encoding assay, wherein each of the plurality of nucleic acid sequences comprises a series of encoder barcode sequences, and wherein each encoder barcode sequence in the given series of encoder barcode sequences corresponds to a conjugate from a set of conjugates that bind to one or more components of a polypeptide in a plurality of polypeptides;

[0389] (b) Generate a binder identifier string for each nucleic acid sequence in a series of encoder barcode sequences, thereby generating multiple binder identifier strings corresponding to multiple nucleic acid sequences;

[0390] (c) Inferring the amino acid sequence of a polypeptide from multiple polypeptides from multiple binder identifier strings based on: (i) the binding characteristics of binders from a set of binders corresponding to binders in the binder identifier strings, and (ii) a calculated probability score of the association between one or more binder identifiers in the binder identifier strings and one or more amino acid sequences of the polypeptides in multiple polypeptides; and

[0391] (d) Based on the calculated probability score and the inferred amino acid sequence, output data relating to at least partial identity and / or quantity of at least one of a variety of polypeptides.

[0392] 2. The computer-implemented method according to Example 1, wherein inferring the amino acid sequence of a polypeptide comprises: (i) for each of a plurality of binder identifier strings, converting a given binder identifier string into one or more peptide reads, the conversion being based on the binding characteristics of binders in a set of binders corresponding to binder identifiers present in the given binder identifier string; and (ii) calculating a probability score for each of the one or more peptide reads, wherein the probability score indicates the probability that a given peptide read generates a given binder identifier string.

[0393] 3. The computer-implemented method according to Example 2, the method further comprising: for each of a plurality of binder identifier strings, filtering out peptide reads generated for one or more peptide reads for a given binder identifier string, the filtering being based on (i) the probability score of each peptide read, and / or (ii) the probability that a given peptide read is generated from an amino acid sequence of a plurality of polypeptides.

[0394] 4. The computer-implemented method according to Example 1, wherein inferring the amino acid sequence of the polypeptide includes:

[0395] (i) For each amino acid sequence in multiple polypeptides, generate one or more simulated binder identifier strings using one or more parameters of the coding assay;

[0396] (ii) For each of one or more simulated binder identifier strings, or for each of one or more amino acid sequences of a polypeptide in a plurality of polypeptides, calculate a probability score based on the probability that a given simulated binder identifier string is associated with one or more amino acid sequences of a polypeptide in a plurality of polypeptides; and

[0397] (iii) Based on the calculated probability scores of one or more simulated combiner identifier strings, match each of the multiple combiner identifier strings with one or more simulated combiner identifier strings.

[0398] 5. The computer-implemented method according to Example 4, wherein one or more parameters of the encoding assay include the efficiency of functionalization of N-terminal amino acid (NTAA) residues of peptides in multiple peptides, the efficiency of cleavage of NTAA residues of peptides in multiple peptides, the efficiency of encoding of NTAA residues of peptides in multiple peptides, or any combination thereof.

[0399] 6. The computer-implemented method according to any one of Examples 1 to 5, wherein inferring the amino acid sequence of the polypeptide comprises: inputting a binder identifier string into a trained machine learning model, wherein the trained machine learning model is trained on empirically determined coding assay data.

[0400] 7. The computer-implemented method according to Example 6, wherein the trained machine learning model is trained using a training dataset comprising one or more isolated peptide samples or their digital representations of binder identifier string data or their digital representations, the isolated peptide samples undergoing a coding assay using the same sample preparation protocol as that used for processing multiple peptides.

[0401] 8. The computer-implemented method according to Example 6, wherein the trained machine learning model is configured to (i) map each of a plurality of binder identifier strings to a specific polypeptide sequence, or (ii) assign a given binder identifier string fractionally to two or more specific polypeptide sequences as part of inferring the amino acid sequence of the polypeptide from the binder identifier.

[0402] 9. The computer-implemented method according to Example 6, wherein the assay parameter data determined empirically includes: (i) the probability of assigning one or more binder identifiers from a set of binders to potential N-terminal amino acid (NTAA) residues of the polypeptide based on the binding characteristics of the binders used in the coding assay; and (ii) the probability of successfully cleaving the N-terminal amino acid residues after encoding them in the coding assay for each potential N-terminal amino acid residue in the polypeptide.

[0403] 10. A computer-implemented method according to any one of Examples 1 to 9, wherein inferring the amino acid sequence of a polypeptide from a binder identifier string comprises: inputting a binder identifier string or a numerical representation thereof into a trained machine learning model, wherein the trained machine learning model is trained on a training dataset comprising a set of simulated binder identifier strings or numerical representations thereof generated for a given input distribution of the polypeptide, the generation being based on assay parameter data determined empirically, the assay parameter data comprising:

[0404] (i) The probability of assigning one or more binder identifiers from a set of binders to potential N-terminal amino acid (NTAA) residues of a polypeptide based on the binding characteristics of the binders used in the coding assay.

[0405] (ii) For each potential NTAA residue in the polypeptide, the probability of cleaving the NTAA residue after encoding the NTAA in the coding assay, and

[0406] (iii) The probability of modifying NTAA residues.

[0407] 11. The computer-implemented method according to any one of Examples 1 to 10, wherein a set of binders comprises at least 5 different binders, and wherein each binder binds one or more components of a polypeptide, each comprising NTAA residues of a polypeptide from a plurality of polypeptides or NTAA residues modified by an N-terminal modifier.

[0408] 12. The computer-implemented method according to any one of Examples 1 to 11, wherein a set of binders comprises at least 5 different binders, and wherein each binder binds one or more components of a polypeptide, each comprising a terminal amino acid or a terminal dipeptide of a polypeptide from a plurality of polypeptides.

[0409] 13. The computer-implemented method according to Example 11 or Example 12, wherein the N-terminal modifier is a compound selected from the group consisting of:

[0410]

[0411] Where M is a metal-bonded group containing sulfonamide, isohydroxamic acid, aminosulfonate, or thioamide; group It is a 5- or 6-membered aromatic ring containing up to three heteroatoms selected from N, O, and S as ring members, and optionally substituted with R; R represents one or two optional substituents selected from the group consisting of: F, Cl, CH3, CF2H, CF3, OH, OCH3, OCF3, NH2, N(CH3)2, NO2, SCH3, SO2CH3, CH2OH, B(OH)2, CN, CONH2, CO2H, CN4H, and CONHCH3; LG is OH, ORQ, or OCC, each RQ is independently aryl or heteroaryl, each optionally substituted with one or more groups selected from halogen, nitro, cyano, sulfonate, carboxylate, and alkylsulfonyl, and the N of the heteroaryl group is optionally oxidized; or RQ can be -C(=O)R or -C(=O)-OR; CC is a cationic counterion; X is one of: O, S, Se, or NH.

[0412] 14. The computer-implemented method according to any one of Examples 1 to 13, wherein a plurality of nucleic acid sequences are generated in a coding assay comprising contacting a plurality of polypeptides with a set of conjugates.

[0413] 15. The computer-implemented method according to any one of Embodiments 1 to 14, wherein a series of encoder barcode sequences comprises at least three different encoder barcode sequences.

[0414] 16. The computer-implemented method according to any one of Embodiments 1 to 15, wherein a series of encoder barcode sequences comprises 4 to 20 different encoder barcode sequences.

[0415] 17. The computer-implemented method according to Example 6, wherein a trained machine learning model is trained on a training dataset generated by performing multiple coding measurements with a predetermined analyte and by inputting a series of encoder barcode sequences or their digital representations and coding measurement parameters generated during each coding measurement into the machine learning model.

[0416] 18. A computer-implemented method according to any one of Examples 1 to 17, wherein a computer model infers two or more amino acid sequences of a polypeptide from a plurality of polypeptides from each binder identifier string, and outputs the probability that a given binder identifier string in the plurality of binder identifier strings originates from one of the two or more amino acid sequences of the polypeptide inferred from the binder identifier string.

[0417] 19. A computer-implemented method according to any one of Embodiments 1 to 18, wherein

[0418] (i) Each of the multiple nucleic acid sequences further includes one or more auxiliary sequences;

[0419] (ii) The method further includes using the one or more processors to assign auxiliary identifiers to at least one of one or more auxiliary sequences in each of a plurality of nucleic acid sequences to generate a plurality of binding sub-identifier strings including the auxiliary identifiers; and

[0420] (iii) One or more auxiliary sequences contain one or more identifier sequences or their complements, one or more spacer sequences or their complements, one or more sequencing primer sequences or their complements, or any combination thereof.

[0421] 20. The computer-implemented method according to Example 19, wherein one or more identifier sequences include one or more sample barcode sequences, one or more bead barcode sequences, one or more unique molecular identifier (UMI) sequences, or any combination thereof.

[0422] 21. The computer-implemented method according to Example 19, wherein the generation of the binder identifier string is based on (i) a specific order of encoder barcode sequences in each of a plurality of nucleic acid sequences, and optionally (ii) a lookup table (LUT) of binder identifiers.

[0423] 22. The computer-implemented method according to embodiment 21, wherein generating the concatenated sub-identifier string further comprises: performing error correction based on an enumeration of error types to account for errors associated with the encoding process used to generate each encoder barcode sequence in a series of encoder barcode sequences.

[0424] 23. The computer-implemented method according to Example 21 or Example 22, wherein generating a binder identifier string includes using a probabilistic model to predict a sequence of binder identifiers for each of a plurality of nucleic acid sequences.

[0425] 24. The computer-implemented method according to any one of Examples 1 to 23, wherein the multiple polypeptides encoded in a plurality of nucleic acid sequences by encoding assay include at least 1,000, at least 10,000, at least 100,000, at least 1,000,000 or more polypeptides.

[0426] 25. The computer-implemented method according to any one of Examples 1 to 24, wherein the multiple polypeptides encoded in a plurality of nucleic acid sequences by encoding assay include at least 500, at least 1000, at least 2000, at least 5000, at least 10,000, at least 20,000 or more different polypeptides.

[0427] 26. The computer-implemented method according to any one of Examples 1 to 25, wherein the binding characteristics of the binder are determined experimentally.

[0428] 27. The computer-implemented method according to any one of Examples 1 to 25, wherein the features are determined computationally.

[0429] 28. The computer-implemented method according to any one of Examples 18 to 26, wherein inferring the amino acid sequence of a polypeptide from a binder identifier string comprises: inputting the binder identifier string and coding assay data into a computer model.

[0430] 29. The computer-implemented method according to Example 28, wherein the computer model is configured to identify a unique binder identifier feature spectrum among a plurality of binder identifier strings as part of inferring the amino acid sequence of a polypeptide from the binder identifier, and wherein a given unique binder identifier feature spectrum comprises a set of binder identifier strings associated with a single polypeptide among a plurality of polypeptides.

[0431] 30. The computer-implemented method according to Example 28, wherein the computer model calculates the binding characteristics of a set of binders based on input coded measurement parameter data.

[0432] 31. The computer-implemented method according to any one of Examples 1 to 30, wherein inferring the amino acid sequence of a polypeptide from a binder identifier comprises: inputting one or more corresponding numeric representations of a binder identifier string and, optionally, one or more corresponding numeric representations encoding assay parameter data into a computer model.

[0433] 32. The computer-implemented method according to any one of Examples 1 to 31, wherein inferring the amino acid sequence of a polypeptide from a binder identifier comprises: outputting one or more corresponding numerical representations of the polypeptide sequence from a computer model.

[0434] 33. The computer-implemented method according to Example 32, wherein the corresponding numerical representation of the binder identifier string is a binder sequence embedding generated from a trained protein language model (pLM), protein basic model, or natural language processing (NLP) model.

[0435] 34. The computer-implemented method according to Example 32, wherein the corresponding numerical representation of the polypeptide sequence indicates a polypeptide sequence embedding generated from a protein language model, a protein basic model, or a natural language processing model.

[0436] 35. The computer-implemented method according to Example 31, wherein the corresponding numerical representation of the coded assay parameter data includes: (i) the concentration of one or more conjugates in a set of conjugates; (ii) the thermodynamic parameters of the conjugates, including the binding rate constant and dissociation rate constant of one or more polypeptide components; (iii) the incubation time of the conjugates; (iv) the washing time of the conjugates after incubation; (v) the ligase concentration; (v) the ligase reaction time; (vi) the thermodynamic parameters of the ligase, including the Michaelis constant and catalytic turnover; (vii) the estimated polypeptide concentration; (viii) the concentration of the lyase; and (ix) the buffer conditions for each step of the coded assay, including the enzyme substrate concentration, salt type, salt concentration, and pH.

[0437] 36. The computer-implemented method according to any one of Examples 1 to 35, wherein the binder identifier string for each nucleic acid sequence is generated by inferring identification information about the corresponding binder for each encoder barcode sequence within a given nucleic acid sequence.

[0438] 37. The computer-implemented method according to Example 22, wherein considering errors associated with the process of generating each encoder barcode sequence in a series of encoder barcode sequences includes determining: (i) a first probability of reading a specified encoder barcode sequence of a specified underlying polypeptide component, and (ii) a second probability of successfully transferring from one polypeptide component to the next polypeptide component during the encoding process.

[0439] 38. The computer-implemented method according to Example 23, wherein the probabilistic model includes a Hidden Markov Model (HMM), and optionally wherein the Hidden Markov Model (HMM) is trained using one or more training datasets comprising a tag pair including a binder identifier and a nucleic acid sequence.

[0440] 39. The computer-implemented method according to Example 38, wherein the Hidden Markov Model (HMM) is trained using an iterative expectation-maximization (EM) algorithm to determine a set of model parameters that maximize the probability of correctly predicting the sequence of the binding identifier of each of a plurality of nucleic acid sequences, and optionally wherein the expectation-maximization (EM) algorithm includes the Baum-Welch algorithm.

[0441] 40. The computer-implemented method according to any one of Examples 1 to 39, wherein one or more components of the bound polypeptide include a post-translational modification of at least one amino acid residue.

[0442] 41. The computer-implemented method according to Example 29, wherein the trained machine learning model is trained using a training dataset comprising binder identifier string data of one or more isolated peptide samples processed using the same sample preparation protocol as that used for processing multiple peptides.

[0443] 42. The computer-implemented method according to Example 41, wherein the training dataset further includes binder identifier string data of background samples processed using the same sample preparation protocol as that used for processing a variety of peptides, and wherein optionally the background samples include plasma samples, urine samples, saliva samples or cell extract samples.

[0444] 43. The computer-implemented method according to any one of Examples 18 to 42, wherein the computer model is further configured to correct the quantity output of at least one polypeptide using a correction factor calculated from a training dataset.

[0445] 44. The computer-implemented method according to Example 8, wherein mapping a binder identifier string from a plurality of binder identifier strings to a specific polypeptide comprises:

[0446] i) Generate a set of k-merged fragments of the combiner identifier string;

[0447] ii) Determine the probability that a given k-mer fragment in a set belongs to a specific polypeptide based on a previously determined probability distribution; and

[0448] iii) Assign a binder identifier string to a specific polypeptide based on a determined probability of a set of k-mer fragments.

[0449] 45. The computer-implemented method according to Example 44, wherein the probability model is trained on measurement parameter data determined empirically.

[0450] 46. ​​The computer-implemented method according to Example 45, wherein for each of a plurality of barcode sequences, the assay parameter data determined empirically includes: the probability of reading a given barcode sequence in a nucleic acid sequence for a given one or more amino acid residues of a corresponding polypeptide bound to the binding moiety.

[0451] 47. The computer-implemented method according to Example 45 or Example 46, wherein for each potential N-terminal amino acid residue in the polypeptide, the empirically determined assay parameter data includes the probability of successfully cleaving the N-terminal amino acid residue during a cycling process for encoding multiple polypeptides in multiple nucleic acid sequences.

[0452] 48. The computer-implemented method according to Example 8, wherein mapping a binder identifier string from a plurality of binder identifier strings to a specific polypeptide comprises: providing the plurality of binder identifier strings as input to a trained machine learning model, wherein the trained machine learning model is configured to assign a given binder identifier string to two or more specific polypeptides in a fractional manner.

[0453] 49. The computer-implemented method according to Example 48, wherein the trained machine learning model is trained on a training dataset comprising a set of simulated binder identifier strings generated for a given input distribution of a peptide, the generation being based on assay parameter data determined empirically, the assay parameter data including:

[0454] i) For each barcode sequence in a plurality of barcode sequences, for a given one or more amino acid residues of the corresponding polypeptide bound to the binding portion, the probability of reading a given barcode sequence in the nucleic acid sequence; and

[0455] ii) For each potential N-terminal amino acid residue in a polypeptide, the probability of successfully cleaving the N-terminal amino acid residue during a cyclic process used to encode multiple polypeptides in multiple nucleic acid sequences.

[0456] 50. The computer-implemented method according to Example 49, wherein a set of simulated combiner identifier strings includes at least 100,000, 500,000, 1M, 5M, 10M, 25M, 50M, 75M, 100M, 200M, 300M, 400M or 500M simulated combiner identifier strings.

[0457] 51. The computer-implemented method according to any one of Examples 1 to 50, wherein the method is further configured to output a confidence interval for partial identity of at least one polypeptide.

[0458] 52. The computer-implemented method according to embodiment 51, wherein the stringency level of the output confidence interval can be selected by the user.

[0459] 53. The computer-implemented method according to Example 52, wherein the stringency level of the confidence interval corresponds to a confidence level of 90%, 95%, 98%, or 99%.

[0460] 54. The computer-implemented method according to Example 6, wherein the trained machine learning model includes a trained artificial neural network model, a trained deep learning model, a trained random forest model, or a trained support vector machine.

[0461] 55. The computer-implemented method according to any one of Examples 1 to 54, the method further comprising performing an iterative expectation maximization (EM) process to refine the quantity output of at least one polypeptide among a plurality of polypeptides.

[0462] 56. The computer-implemented method according to Example 55, wherein the iterative EM process repeatedly includes:

[0463] i) Based on the current estimate of which peptides are present in a variety of peptides, find the optimal allocation of the binding moiety to each barcode sequence in the nucleic acid sequence;

[0464] ii) Based on the optimal allocation of the binding moiety to each different barcode sequence in the nucleic acid sequence, find the best updated estimate of which peptides are present in a variety of peptides; and

[0465] iii) Based on the best updated estimate of which peptides are present in a variety of peptides, determine the amount of at least one peptide present in the variety of peptides.

[0466] 57. The computer-implemented method according to Example 56, wherein steps (i) to (iii) are repeated until the difference between the amount of at least one polypeptide determined in one iteration and the next is less than a specified threshold, or until a specified maximum number of iterations has been reached.

[0467] 58. The computer-implemented method according to any one of Examples 1 to 57, the method further comprising determining a plurality of nucleic acid sequences, optionally wherein the nucleic acid sequences are determined by DNA sequencing.

[0468] 59. The computer-implemented method according to any one of Examples 1 to 58, the method further comprising determining the amino acid sequence of each of a plurality of polypeptides.

[0469] 60. The computer-implemented method according to any one of Examples 1 to 59, wherein the plurality of nucleic acid sequences comprises at least 10,000, 100,000 or 1M nucleic acid sequences.

[0470] 61. The computer-implemented method according to any one of Examples 1 to 60, wherein the plurality of polypeptides includes at least 100, 500, 1,000, 2,000, 5,000 or 10,000 different polypeptides.

[0471] 62. The computer-implemented method according to Example 2, wherein at least two peptide reads are generated for at least one nucleic acid sequence among a plurality of nucleic acid sequences.

[0472] 63. A computer-implemented method comprising:

[0473] a) Receiving amino acid sequences of multiple peptides at one or more processors;

[0474] b) Using the one or more processors, generate multiple peptide sequences based on the amino acid sequence;

[0475] c) Based on the sequence of amino acid residues in the peptide sequence and the binding characteristics of multiple binding moieties, one or more processors are used to convert each peptide sequence in the multiple peptide sequences into a binder identifier string, wherein each of the multiple binding moieties binds to a component of the polypeptide and corresponds to a binder identifier, and wherein the binder identifier corresponds to an encoder barcode sequence; and

[0476] d) Using one or more processors, multiple binder identifier strings are fed as input to a trained model, wherein the trained model is configured to convert the binder identifier strings into nucleic acid sequences based on a set of probabilities of multiple binding moieties binding to components of multiple peptides and output multiple nucleic acid sequences corresponding to multiple peptides.

[0477] 64. The computer-implemented method according to Example 63, further comprising comparing a plurality of nucleic acid sequences output by a trained model with a plurality of nucleic acid sequences determined by subjecting a plurality of polypeptides to coding assays and sequencing the resulting extended recording tags.

[0478] 65. A system comprising:

[0479] One or more processors; and

[0480] A memory, communicatively coupled to one or more processors and configured to store instructions that, when executed by one or more processors, cause the system to perform a computer-implemented method according to any one of embodiments 1 to 64.

[0481] 66. A non-transitory computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by one or more processors of a system, cause the system to perform a computer-implemented method according to any one of embodiments 1 to 64.

[0482] Example

[0483] The following examples are provided to illustrate, but do not limit, the methods, compositions, and uses provided herein. Some aspects of the present invention, including but not limited to embodiments of peptide sequencing assays, methods for information transfer between encoded tags and recording tags, methods for modifying (functionalizing) the terminal amino acid residues of peptides, methods for generating specific binders that recognize the modified terminal amino acid residues of peptides, methods for cleaving the modified terminal amino acid residues of peptides, methods for manufacturing nucleotide-peptide conjugates, methods for attaching nucleotide-peptide conjugates to a support, methods for generating barcodes, and methods for analyzing extended recording tags, are disclosed in earlier publications, such as US 2019 / 0145982 A1, US 2020 / 0348308A1, US 2020 / 0348307 A1, US 2021 / 0208150 A1, US 2022 / 0049246 A1, US 2022 / 0283175A1, US 2022 / 0144885 A1, US 2022 / 0227889 A1, US The contents of US 2021 / 0214701 A1, US 2024 / 0294981A1, US 2024 / 0053350 A1 and US 2023 / 0136966 A1 are incorporated herein by reference in their entirety.

[0484] Example 1 – Coding Determination Method

[0485] The bioinformatics tools described herein complement the recently described methods for peptide analysis employing nucleic acid-based peptide encoding techniques (i.e., reverse translation assays).

[0486] Figure 6 An exemplary design of a solid support (such as agarose beads) configured for peptide analysis / reverse translation assays is depicted. The beads comprise multiple covalently attached peptides, each associated with a barcoded nucleic acid recording tag (RT). Each bead uses approximately 10^7 RTs. Two barcodes are shown: a BBC – a bead barcode enabling sample multiplexing (each peptide sample will be incubated with a set of beads sharing a unique bead barcode); and a PBC – a peptide barcode that uniquely identifies the attached peptide (also known as an RT barcode) or the fraction or compartment from which the attached peptide originates. An ILM is used for peptide attachment and amplification of the extended RT generated during encoding. A spacer A (Sp_A) is required for the transfer (encoding) of information, forming complementary regions between the RT and the coding tag (CT) of the binding reagent (which binds to the peptide associated with the RT), and enabling connections between the RT and CT.

[0487] Figure 7A and Figure 7B Two exemplary peptide assays involving information transfer are described. Figure 7A and Figure 7B The left image shows a polypeptide attached to a solid support and associated with a barcoded nucleic acid recording tag (RT), similar to... Figure 6 As shown in the diagram. Next, the cycle of encoding (information transfer) is illustrated, which includes three steps: binding, ligation, and digestion with restriction enzymes (RE). See [link to diagram]. Figure 7A ) or 4 steps (binding, ligation, extension, and digestion with restriction enzymes (RE), see Figure 7B During the binding step, the binding agent binds to the peptide (e.g., a modified NTAA of the peptide), bringing the coded tag (CT) attached to the binding agent close to the RT. In some embodiments, the binding agent may attach or connect to the coded tag at locations other than those shown (e.g., in the loop region of the CT or elsewhere). The binding agent, as shown, attaches to the CT via a linker. The coded tag may contain a binding agent-specific barcode (SB), a 2 bp spacer, and an IIS-type restriction enzyme site (RS). Figure 7A The embodiment utilizes only double-chain connections during information transfer (connecting both chains of the double-chain region or connection of the CT to the double-chain region or connection of the RT), while Figure 7B The embodiments utilize both on-chain linkage (linking a single strand of the double-stranded region of CT to the double-stranded region of RT; also referred to as "single-strand linkage") and extension (polymerase extension of the 3' (non-linked) end of RT). Both embodiments utilize a digestion step (cleaving the double-stranded extended recording tag with a nucleic acid lysis reagent (such as RE), which enables the release of binding reagents and generates ends of the double-stranded extended recording tag suitable for further coding cycles (making the ends of the extended RT compatible with the ends of new coding tags for the next binding reagent). Suitable ends of the extended RT may include short single-stranded overhangs or may be blunt ends. Next, in some embodiments, after the information transfer cycle, a portion of the peptide for analysis may be removed from the peptide (e.g., one or two terminal amino acid residues). The binding / linking / digestion cycle may be repeated one or more times with additional binding reagents and coding tags to further extend the recording tag associated with the peptide. Thus, multiple binding events of the peptide can be encoded into the extended recording tag.

[0488] Example 2 – Decoding encoded peptide information

[0489] Figure 8 An exemplary sequence read (SEQ ID NO:12) corresponding to the extended RT obtained by the encoding method is depicted (see, for example, Figure 2 or Figure 7A and Figure 7BThe method described in [the document] is described above. The extended RT is constructed prior to the assay and includes two separate barcodes (BBC and RT barcodes) located in the original (non-extended) RT, followed by one or more spacer-CT barcode pairs, each of which is added to the RT or extended RT during a single coding cycle (the CT barcode is transferred from the CT of the binding reagent to the peptide to the RT during the current coding cycle), and the additional spacer-CT barcode pairs are generated from additional coding cycles. Finally, the extended RT further includes an LM sequence (for extended RT amplification) and a CBC barcode attached during the final capping step, representing the end of the sequence read. The algorithm is used to generate barcode-level alignments of nucleic acid sequence reads based on the illustrated lookup table and considering error correction. In this example, only two binding reagents are used, and only two corresponding CT barcodes (SB04 and SB05) exist in the lookup table. The computer model generates one or more identifier strings, including a binder identifier and auxiliary identifiers, as output for each analyzed sequence read.

[0490] Figure 9 Provides the generation of reads for each analyzed nucleic acid sequence (e.g.) Figure 8 This is a non-limiting example of the output of a computer model of one or more binder identifier strings (as described in [the document]). For example, the first line represents the nucleic acid sequence of a sequence read or a portion thereof, and the second line represents the corresponding binder identifier that constitutes the identifier string corresponding to the sequence read. BBC = beaded barcode. UMI8.0 = unique molecular identifier (8 nucleotides in length). SB49 = RT barcode. GC = GC spacer. SB04 = CT barcode. AG = AG spacer. LM1.0 = amplification primer sequence. CBC = end-capping barcode.

[0491] Example 3 – Proteomics Analysis System

[0492] Figure 10 An exemplary schematic diagram of a proteomics analysis system according to one embodiment described herein is provided, which enables high-throughput analysis of peptides (e.g., peptides and peptides or fragments thereof) based on amino acid sequence data encoded in the form of barcoded nucleic acid sequences. Figure 10 An exemplary flowchart illustrates how identifier strings undergo preprocessing for peptide-level identification and quantification before being processed using a peptide alignment procedure (“top-down” algorithm). The methods described herein are used to analyze sequence read data obtained by sequencing barcoded nucleic acid sequences and are capable of correcting short (e.g., 10-20 amino acid sequence fragments corresponding to peptides) and error-prone sequence reads (e.g., including substitutions, deletions, and phase shifts) that are generally unsuitable for traditional “seed and extension” methods of sequence alignment to identify the peptide or polypeptide from which the reads originate.

[0493] The disclosed method is based on several underlying concepts, such as the ability to pre-calculate the frequency of high-probability proteomic coding error events (e.g., errors in converting amino acid sequences into nucleic acid sequences, and errors in correctly transferring from one amino acid residue to the next during the coding process), and to correct these events using empirical error models. Therefore, the disclosed method provides direct detection and quantification of peptides and polypeptides, rather than relying on alignment with a reference sequence.

[0494] like Figure 10 As shown, an index of the given proteome of interest is prepared as part of the preprocessing steps. Reference proteome data for the proteome of interest (e.g., for a given organism) is obtained (e.g., experimentally obtained and / or downloaded from a suitable proteomics database; peptide abundance can be obtained, for example, from PaxDb (see, e.g., Huang et al. (2023), “PaxDb 5.0: Curated Polypeptide Quantification Data Suggests Adaptive Proteome Changes”, Biorxiv, doi: 10.1101 / 2023.04.16.536357v1), and peptide sequences can be obtained, for example, from UniProt (see, e.g., The UniProt Consortium (2021), “UniProt: the universal polypeptide knowledgebase in 2021”, Nucleic Acids Research 49(D1): (D480–D489) and processed using an indexer to generate an index of the proteome capable of rapid sequence lookup. In some cases, the indexer can generate the index by, for example, by simulating fragmentation of peptide sequences in a reference proteome in a computer and performing sample preparation and coding assay steps to generate a lookup table of short binder identifier strings and the probabilities of their generation from corresponding amino acid sequences of various different peptides derived from the proteome (see, for example, Example 8 below). In some embodiments, the lookup table includes abbreviated codes, each reflecting the specificity of the binder for a particular terminal amino acid residue. In some embodiments, the lookup table includes short nucleic acid sequences instead of short binder identifier strings.

[0495] Figure 11An exemplary schematic diagram of potential errors that can occur when encoding peptide sequence information as barcoded nucleic acid sequences is provided. This illustrates a probabilistic model consisting of transitions between peptide states in subsequent coding cycles (upper layer) and emission within a single coding cycle, based on simulated predictions of the encoded output using terminal amino acid (TAA)-specific binders. The errors shown include transition errors (and associated probabilities) and emission errors (and associated probabilities). Transition errors include errors (and associated probabilities) in the coding assay continuing from one coding cycle to another (e.g., from one amino acid residue to the next), such as errors caused by the failure of the terminal amino acid (TAA) residue of the peptide analyzed after a coding event to cleave to generate a newly formed terminal amino acid residue. Figure 11 In the middle, the current TAA of polypeptide (G) has a certain probability (in Figure 11 In a non-limiting ...

Claims

1. A computer-implemented method for analyzing multiple polypeptides encoded in multiple nucleic acid sequences by a coding assay, the method comprising: (a) Receiving at one or more processors the plurality of nucleic acid sequences generated from the coding assay, wherein each of the plurality of nucleic acid sequences comprises a series of encoder barcode sequences, and wherein each encoder barcode sequence in the given series of encoder barcode sequences corresponds to a conjugate from a set of conjugates that bind to one or more components of the polypeptide in the plurality of polypeptides; (b) Generate a binder identifier string for each of the plurality of nucleic acid sequences based on a series of encoder barcode sequences, thereby generating a plurality of binder identifier strings corresponding to the plurality of nucleic acid sequences; (c) Inferring the amino acid sequence of the polypeptide in the plurality of polypeptides from the binder identifiers in the binder identifier strings in the plurality of binder identifier strings based on: (i) the binding characteristics of the binder from the set of binders corresponding to the binder identifiers in the binder identifier strings, and (ii) a probability score of the calculated association between one or more binder identifiers in the binder identifier strings and one or more amino acid sequences of the polypeptide in the plurality of polypeptides. as well as (d) Based on the calculated probability score and the inferred amino acid sequence, output data relating to at least partial identity and / or quantity of at least one of the multiple polypeptides.

2. The computer-implemented method according to claim 1, wherein inferring the amino acid sequence of the polypeptide comprises: (i) For each of the plurality of binder identifier strings, convert the given binder identifier string into one or more peptide reads based on the binding feature of the binder corresponding to the set of binders present in the given binder identifier string, and (ii) calculate a probability score for each of the one or more peptide reads, wherein the probability score indicates the probability that a given peptide read generates a given binder identifier string.

3. The computer-implemented method according to claim 2, further comprising: For each of the plurality of binder identifier strings, peptide reads generated for a given binder identifier string are filtered out based on: (i) the probability score of each peptide read, and / or (ii) the probability that a given peptide read is generated from the amino acid sequence of the plurality of polypeptides.

4. The computer-implemented method according to claim 1, wherein inferring the amino acid sequence of the polypeptide comprises: (i) For each amino acid sequence in the plurality of polypeptides, one or more simulated binder identifier strings are generated using one or more parameters of the coding assay; (ii) For each of the one or more simulated binder identifier strings, or for each of the one or more amino acid sequences of the polypeptides in the plurality of polypeptides, calculate a probability score based on the probability that a given simulated binder identifier string is associated with one or more amino acid sequences of the polypeptides in the plurality of polypeptides. as well as (iii) Based on the calculated probability score of the one or more simulated combiner identifier strings, match each of the plurality of combiner identifier strings with the one or more simulated combiner identifier strings.

5. The computer-implemented method according to claim 4, wherein the one or more parameters of the encoding determination method include the efficiency of functionalization of N-terminal amino acid (NTAA) residues of the polypeptides in the plurality of polypeptides, the efficiency of cleavage of NTAA residues of the polypeptides in the plurality of polypeptides, the efficiency of encoding of NTAA residues of the polypeptides in the plurality of polypeptides, or any combination thereof.

6. The computer-implemented method according to any one of claims 1 to 5, wherein inferring the amino acid sequence of the polypeptide comprises: The conjugate identifier string is input into a trained machine learning model, wherein the trained machine learning model is trained on empirically determined coding assay data.

7. The computer-implemented method of claim 6, wherein the trained machine learning model is trained using a training dataset comprising one or more isolated peptide samples or their digital representations of binder identifier string data or their digital representations, the isolated peptide samples undergoing the coding assay using the same sample preparation protocol as used for processing the plurality of peptides.

8. The computer-implemented method of claim 6, wherein the trained machine learning model is configured to (i) map each of the plurality of binder identifier strings to a specific polypeptide sequence, or (ii) assign a given binder identifier string fractionally to two or more specific polypeptide sequences as part of inferring the amino acid sequence of the polypeptide from the binder identifier.

9. The computer-implemented method according to claim 6, wherein the measurement parameter data determined empirically includes: (i) the probability of assigning one or more binder identifiers from the set of binders to potential N-terminal amino acid (NTAA) residues of the polypeptide based on the binding characteristics of the binders used in the coding assay; and (ii) the probability of successfully cleaving the N-terminal amino acid residue after encoding it in the coding assay for each potential N-terminal amino acid residue in the polypeptide.

10. The computer-implemented method according to any one of claims 1 to 9, wherein inferring the amino acid sequence of the polypeptide from the binder identifier string comprises: The binder identifier string or its numerical representation is input into a trained machine learning model, wherein the trained machine learning model is trained on a training dataset comprising a set of simulated binder identifier strings or their numerical representations generated based on empirically determined assay parameter data for a given input distribution of the peptide, the empirically determined assay parameter data including: (i) The probability of assigning one or more binder identifiers from the set of binders to potential N-terminal amino acid (NTAA) residues of a polypeptide based on the binding characteristics of the binders used in the coding assay; (ii) For each potential NTAA residue in the polypeptide, the probability of cleaving the NTAA residue after it has been encoded in the coding assay, and (iii) The probability of modifying the NTAA residue.

11. The computer-implemented method according to any one of claims 1 to 10, wherein the set of binders comprises at least five different binders, and wherein each binder binds one or more components of the polypeptide that each of the plurality of polypeptides contains NTAA residues of the polypeptide or NTAA residues modified by an N-terminal modifier.

12. The computer-implemented method according to any one of claims 1 to 11, wherein the set of binders comprises at least five different binders, and wherein each binder binds one or more components of the polypeptide that each of the plurality of polypeptides contains a terminal amino acid or a terminal dipeptide of the polypeptide.

13. The computer-implemented method according to claim 11 or claim 12, wherein the N-terminal modifier is a compound selected from the group consisting of: Where M is a metal-bonded group containing sulfonamide, isohydroxamic acid, aminosulfonate, or thioamide; group It is a 5- or 6-membered aromatic ring containing up to three heteroatoms selected from N, O, and S as ring members, and optionally substituted with R; R represents one or two optional substituents selected from the group consisting of: F, Cl, CH3, CF2H, CF3, OH, OCH3, OCF3, NH2, N(CH3)2, NO2, SCH3, SO2CH3, CH2OH, B(OH)2, CN, CONH2, CO2H, CN4H, and CONHCH3; LG is OH, ORQ, or OCC, each RQ is independently aryl or heteroaryl, each optionally substituted with one or more groups selected from halogen, nitro, cyano, sulfonate, carboxylate, and alkylsulfonyl, and the N of the heteroaryl group is optionally oxidized; or RQ can be -C(=O)R or -C(=O)-OR; CC is a cationic counterion; X is one of: O, S, Se, or NH.

14. The computer-implemented method according to any one of claims 1 to 13, wherein the plurality of nucleic acid sequences are generated in a coding assay, the coding assay comprising contacting the plurality of polypeptides with the set of binders.

15. The computer-implemented method according to any one of claims 1 to 14, wherein the series of encoder barcode sequences comprises at least three different encoder barcode sequences.

16. The computer-implemented method according to any one of claims 1 to 15, wherein the series of encoder barcode sequences comprises 4 to 20 different encoder barcode sequences.

17. The computer-implemented method of claim 6, wherein the trained machine learning model is trained on a training dataset, the training dataset being generated by performing multiple coding assays with a predetermined analyte and by inputting the series of encoder barcode sequences or their digital representations and coding assay parameters generated during each coding assay into the machine learning model.

18. The computer-implemented method according to any one of claims 1 to 17, wherein the computer model infers two or more amino acid sequences of a polypeptide among the plurality of polypeptides from each binder identifier string, and outputs the probability that a given binder identifier string among the plurality of binder identifier strings originates from one of the two or more amino acid sequences of the polypeptide inferred from the binder identifier string.

19. The computer-implemented method according to any one of claims 1 to 18, wherein: (i) Each of the plurality of nucleic acid sequences further includes one or more auxiliary sequences; (ii) The method further includes using the one or more processors to assign auxiliary identifiers to at least one of the one or more auxiliary sequences in each of the plurality of nucleic acid sequences to generate a plurality of binding sub-identifier strings including the auxiliary identifiers; as well as (iii) The one or more auxiliary sequences comprise one or more identifier sequences or their complements, one or more spacer sequences or their complements, one or more sequencing primer sequences or their complements, or any combination thereof.

20. The computer-implemented method of claim 19, wherein the one or more identifier sequences comprise one or more sample barcode sequences, one or more bead barcode sequences, one or more unique molecular identifier (UMI) sequences, or any combination thereof.

21. The computer-implemented method of claim 19, wherein the generation of the binder identifier string is based on (i) a specific order of the encoder barcode sequence in each of the plurality of nucleic acid sequences, and optionally (ii) a lookup table (LUT) for the binder identifier.

22. The computer-implemented method of claim 21, wherein generating the concatenation identifier string further comprises: Error correction is performed based on an enumeration of error types to account for errors associated with the encoding process used to generate each encoder barcode sequence in the series of encoder barcode sequences.

23. The computer-implemented method of claim 21 or claim 22, wherein generating the binder identifier string comprises using a probabilistic model to predict a sequence of binder identifiers for each of the plurality of nucleic acid sequences.

24. The computer-implemented method according to any one of claims 1 to 23, wherein the plurality of polypeptides encoded in the plurality of nucleic acid sequences by the coding assay comprises at least 1,000, at least 10,000, at least 100,000, at least 1,000,000 or more polypeptides.

25. The computer-implemented method according to any one of claims 1 to 24, wherein the plurality of polypeptides encoded in the plurality of nucleic acid sequences by the coding assay comprises at least 500, at least 1000, at least 2000, at least 5000, at least 10,000, at least 20,000 or more different polypeptides.

26. The computer-implemented method according to any one of claims 1 to 25, wherein the binding characteristics of the combiner are determined experimentally.

27. The computer-implemented method according to any one of claims 1 to 25, wherein the combination feature is determined computationally.

28. The computer-implemented method according to any one of claims 18 to 26, wherein inferring the amino acid sequence of the polypeptide from the binder identifier string comprises: The combined sub-identifier string and the encoding determination data are input into the computer model.

29. The computer-implemented method of claim 28, wherein the computer model is configured to identify a unique binder identifier feature among the plurality of binder identifier strings as part of inferring the amino acid sequence of a polypeptide from the binder identifier, and wherein a given unique binder identifier feature comprises a set of binder identifier strings associated with a single polypeptide among the plurality of polypeptides.

30. The computer-implemented method of claim 28, wherein the computer model calculates the binding characteristics of the binders in the set of binders based on the input coded measurement parameter data.

31. The computer-implemented method according to any one of claims 1 to 30, wherein inferring the amino acid sequence of the polypeptide from the binder identifier comprises: One or more corresponding numeric representations of the combined sub-identifier string and one or more corresponding numeric representations of the optionally encoded measurement parameter data are input into the computer model.

32. The computer-implemented method according to any one of claims 1 to 31, wherein inferring the amino acid sequence of the polypeptide from the binder identifier comprises: Output one or more corresponding numerical representations of the polypeptide sequence from a computer model.

33. The computer-implemented method of claim 32, wherein the corresponding number of the binder identifier string represents a binder sequence embedding generated from a trained protein language model (pLM), protein basic model, or natural language processing (NLP) model.

34. The computer-implemented method of claim 32, wherein the corresponding number of the polypeptide sequence represents a polypeptide sequence embedding generated from a protein language model, a protein basic model, or a natural language processing model.

35. The computer-implemented method according to claim 31, wherein the corresponding numerical representation of the coded measurement parameter data includes: (i) the concentration of one or more conjugates in the set of conjugates; (ii) the thermodynamic parameters of the conjugates, including the binding rate constant and dissociation rate constant of one or more polypeptide components; (iii) the incubation time of the conjugates; (iv) the washing time of the conjugates after incubation; (v) the ligase concentration; (v) the ligase reaction time; (vi) the thermodynamic parameters of the ligase, including the Michaelis constant and catalytic turnover; (vii) the estimated polypeptide concentration; (viii) the concentration of the lyase; and (ix) the buffer conditions for each step of the coded assay, including the enzyme substrate concentration, salt type, salt concentration, and pH.

36. The computer-implemented method according to any one of claims 1 to 35, wherein the binder identifier string for each nucleic acid sequence is generated by inferring identification information of the corresponding binder for each encoder barcode sequence within a given nucleic acid sequence.

37. The computer-implemented method of claim 22, wherein considering the error associated with the process for generating each encoder barcode sequence in the series of encoder barcode sequences includes determining: (i) a first probability of reading a specified encoder barcode sequence of a specified underlying polypeptide component, and (ii) a second probability of successfully transferring from one polypeptide component to the next polypeptide component during the encoding process.

38. The computer-implemented method of claim 23, wherein the probabilistic model comprises a Hidden Markov Model (HMM), and optionally wherein the Hidden Markov Model (HMM) is trained using one or more training datasets comprising a tag pair comprising a binder identifier and a nucleic acid sequence.

39. The computer-implemented method of claim 38, wherein the Hidden Markov Model (HMM) is trained using an iterative expectation-maximization (EM) algorithm to determine a set of model parameters that maximizes the probability of correctly predicting the sequence of the binding identifier of each of the plurality of nucleic acid sequences, and optionally wherein the expectation-maximization (EM) algorithm includes the Baum-Welch algorithm.

40. The computer-implemented method according to any one of claims 1 to 39, wherein one or more components of the polypeptide bound to the binding portion comprise a post-translational modification of at least one amino acid residue.

41. The computer-implemented method of claim 29, wherein the trained machine learning model is trained using a training dataset comprising binder identifier string data of one or more isolated peptide samples, the one or more isolated peptide samples being processed using the same sample preparation protocol as used for processing the plurality of peptides.

42. The computer-implemented method of claim 41, wherein the training dataset further comprises binder identifier string data of background samples processed using the same sample preparation protocol as that used for processing the plurality of peptides, and wherein optionally the background samples include plasma samples, urine samples, saliva samples, or cell extract samples.

43. The computer-implemented method according to any one of claims 18 to 42, wherein the computer model is further configured to correct the quantity output of the at least one polypeptide using a correction factor calculated from a training dataset.

44. The computer-implemented method of claim 8, wherein mapping a binder identifier string among the plurality of binder identifier strings to a specific polypeptide comprises: i) Generate a set of k-merged fragments of the conjugate identifier string; ii) Determine the probability that a given k-mer fragment in the set belongs to a specific polypeptide based on a previously determined probability distribution; as well as iii) Assign the binder identifier string to the specific polypeptide based on the determined probability of the set of k-mer fragments.

45. The computer-implemented method of claim 44, wherein the probability model is trained on measurement parameter data determined empirically.

46. ​​The computer-implemented method of claim 45, wherein the empirically determined measurement parameter data includes: For each barcode sequence in a plurality of barcode sequences, for a given one or more amino acid residues of the corresponding polypeptide bound to the binding portion, the probability of reading a given barcode sequence in the nucleic acid sequence.

47. The computer-implemented method according to claim 45 or claim 46, wherein the empirically determined measurement parameter data includes: For each potential N-terminal amino acid residue in a polypeptide, the probability of successfully cleaving the N-terminal amino acid residue during a cyclic process used to encode the various polypeptides in the plurality of nucleic acid sequences.

48. The computer-implemented method of claim 8, wherein mapping a binder identifier string among the plurality of binder identifier strings to a specific polypeptide comprises: The plurality of binder identifier strings are provided as input to the trained machine learning model, wherein the trained machine learning model is configured to assign a given binder identifier string to two or more specific peptides in a fractional manner.

49. The computer-implemented method of claim 48, wherein the trained machine learning model is trained on a training dataset comprising a set of simulated binder identifier strings generated for a given input distribution of a peptide based on empirically determined assay parameter data, wherein the empirically determined assay parameter data includes: i) For each barcode sequence in a plurality of barcode sequences, for a given one or more amino acid residues of the corresponding polypeptide bound to the binding portion, the probability of reading a given barcode sequence in the nucleic acid sequence; as well as ii) For each potential N-terminal amino acid residue in a polypeptide, the probability of successfully cleaving the N-terminal amino acid residue during a cyclic process for encoding the multiple polypeptides in the multiple nucleic acid sequences.

50. The computer-implemented method of claim 49, wherein the set of simulated combiner identifier strings comprises at least 100,000, 500,000, 1M, 5M, 10M, 25M, 50M, 75M, 100M, 200M, 300M, 400M, or 500M simulated combiner identifier strings.

51. The computer-implemented method according to any one of claims 1 to 50, wherein the method is further configured to output a confidence interval for the partial identity of the at least one polypeptide.

52. The computer-implemented method of claim 51, wherein the stringency level of the output confidence interval can be selected by the user.

53. The computer-implemented method of claim 52, wherein the stringency level of the confidence interval corresponds to a confidence level of 90%, 95%, 98%, or 99%.

54. The computer-implemented method of claim 6, wherein the trained machine learning model comprises a trained artificial neural network model, a trained deep learning model, a trained random forest model, or a trained support vector machine.

55. The computer-implemented method according to any one of claims 1 to 54, the method further comprising performing an iterative expectation-maximization (EM) process to refine the quantity output of at least one of the plurality of polypeptides.

56. The computer-implemented method of claim 55, wherein the iterative EM process repeatedly comprises: i) Based on the current estimate of which polypeptides are present among the various polypeptides, find the optimal allocation of the binding moiety to each barcode sequence in the nucleic acid sequence; ii) Based on the optimal allocation of the binding portion to each different barcode sequence in the nucleic acid sequence, find an updated optimal estimate of which polypeptides are present among the multiple polypeptides; and iii) Based on the updated best estimate of which polypeptides are present among the plurality of polypeptides, determine the amount of at least one polypeptide present among the plurality of polypeptides.

57. The computer-implemented method of claim 56, wherein steps (i) to (iii) are repeated until the difference between the amount of the at least one polypeptide determined in one iteration and the next is less than a specified threshold, or until a specified maximum number of iterations has been reached.

58. The computer-implemented method according to any one of claims 1 to 57, further comprising determining the nucleic acid sequences of the plurality of nucleic acid sequences, optionally wherein the nucleic acid sequences are determined by DNA sequencing.

59. The computer-implemented method according to any one of claims 1 to 58, further comprising determining the amino acid sequence of each of the plurality of polypeptides.

60. The computer-implemented method according to any one of claims 1 to 59, wherein the plurality of nucleic acid sequences comprises at least 10,000, 100,000, or 1M nucleic acid sequences.

61. The computer-implemented method according to any one of claims 1 to 60, wherein the plurality of polypeptides comprises at least 100, 500, 1,000, 2,000, 5,000 or 10,000 different polypeptides.

62. The computer-implemented method of claim 2, wherein at least two peptide reads are generated for at least one of the plurality of nucleic acid sequences.

63. A computer-implemented method comprising: a) Receiving amino acid sequences of multiple peptides at one or more processors; b) Using the one or more processors, generate multiple peptide sequences based on the amino acid sequence; c) Based on the sequence of amino acid residues in the peptide sequence and the binding characteristics of multiple binding moieties, the one or more processors are used to convert each peptide sequence in the plurality of peptide sequences into a binder identifier string, wherein each of the plurality of binding moieties binds to a component of the polypeptide and corresponds to a binder identifier, and wherein the binder identifier corresponds to an encoder barcode sequence. as well as d) Using the one or more processors, a plurality of binder identifier strings are provided as input to a trained model, wherein the trained model is configured to convert the binder identifier strings into nucleic acid sequences based on a set of probabilities of the plurality of binding portions binding to components of the plurality of polypeptides and output a plurality of nucleic acid sequences corresponding to the plurality of polypeptides.

64. The computer-implemented method of claim 63, further comprising comparing the plurality of nucleic acid sequences output by the trained model with a plurality of nucleic acid sequences determined by subjecting the plurality of polypeptides to coding assays and sequencing the resulting extended recording tags.

65. A system comprising: One or more processors; as well as A memory, communicatively coupled to the one or more processors and configured to store instructions that, when executed by the one or more processors, cause the system to perform a computer-implemented method according to any one of claims 1 to 64.

66. A non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions that, when executed by one or more processors of a system, cause the system to perform a computer-implemented method according to any one of claims 1 to 64.

Images