Immunological entity sequence data processing

Abstract

Immunological entity sequences are clustered by a two-step process based on their similarity, i.e., their ability to react to the same antigen. In the first step, the sequences are converted into numerical vectors, and these numerical vectors are clustered into initial clusters. In the second step, the initial clusters are refined into final clusters, which can be done using the sequence data itself rather than numerical vectors. Each final cluster may be assigned a representative sequence to reduce the dimensionality of the cluster.

Description

Technical Field

[0001] The present invention relates to methods and computer programs for processing sequence data from immunological entities such as T cell receptor (TCR) or B cell receptor (BCR) sequence data.

Background Art

[0002] The T cell receptor (TCR) is a protein complex found on the surface of T cells that is responsible for the recognition of fragments of antigenic peptides. The binding between TCR and antigenic peptides is of relatively low affinity and is degenerate. Many TCRs recognize the same antigenic peptide, and many antigenic peptides are recognized by the same TCR. Therefore, it is desirable to be able to identify similar TCR sequences, for example, based on their ability to react to the same or similar antigenic peptides.

[0003] The TCR protein is composed of two chains called the alpha chain and the beta chain, and these chains are held together by disulfide bonds. Each chain contains a variable (V) region, a diversity (D) region, and a joining (J) region, and these regions are responsible for antigen recognition. The V regions of the alpha and beta chains together form an antigen-binding site or complementarity-determining region (CDR), which is responsible for specifically recognizing the antigen.

[0004] The diversity of the TCR structure is enabled by the diversity of the amino acids that make up the V, D, and J regions of the alpha and beta chains. These regions contain a high variability of amino acid sequences, which allows for a wide range of antigen recognition. The amino acids in the CDR of the TCR chain form a three-dimensional structure that enables the TCR to specifically recognize and bind to the antigen.

[0005] The sequence encoded by the V(D)J junction is called complementarity-determining region 3 or CDR3. This sequence has the highest variability in both the alpha and beta chains and determines the ability of T cells to recognize antigenic peptides.

[0006] Different amino acids possess distinct properties such as size, charge, and polarity, which play a crucial role in determining the structure and function of the transcatheter cellular network (TCR). For example, amino acids with large side chains, such as arginine or lysine, may be involved in the formation of salt bridges or hydrogen bonds, while amino acids with small side chains, such as glycine or alanine, may contribute to the flexibility of the TCR structure.

[0007] Non-patent document 1 summarizes the repertory sequencing technologies available in 2012 for the study of B cell and T cell repertoires. Next-generation immunosequencing technologies (such as the OmniScope T Cell Sequencing Platform OS-T) are capable of sequencing up to 1 million T cells per sample and characterizing immune responses with sensitivity that enables the early and high-resolution detection of repertory changes for advanced translational and clinical research. However, improved technologies are needed to process and analyze the resulting large amounts of sequencing data. [Prior art documents] [Non-patent literature]

[0008] [Non-Patent Document 1] “Rep-Seq:uncovering the immunological repertoire through next-generation sequencing”, Benichou et al., Immunology.2012 Mar;135(3):183-191 [Non-Patent Document 2] “Modeling aspects of the language of life through transfer-learning protein sequences” Heinzinger et al.BMC Bioinformatics (2019)20:723 [Non-Patent Document 3] "Amino acid substitution matrix from protein blocks" Henikoff S, Henikoff JG,.Proc Natl Acad Sci USA 89:10915-10919(1992) [Non-Patent Document 4] "Estimation of effective interresidue contact energies from protein crystal structures:quasi-chemical approximation", Miyazawa S, Jernigan R, Macromolecules 1985,18,3,534-552. [Non-Patent Document 5] "Density-Based Clustering Based on Hierarchical Density Estimates", Campello, RJGB, Moulavi, D., Sander, J. (2013. In: Pei, J., Tseng, VS, Cao, L., Motoda, H., Xu, G. (eds) Advances in Knowledge Discovery and Data Mining. PAKDD 2013. Lecture Notes in Computer Science(),vol 7819.Springer,Berlin,Heidelberg [Non-Patent Document 6] "Binary Codes Capable of Correcting Deletions, Insertions and Reversals";Levenshtein V,Soviet Physics-Doklady,Vo.10,no.8,February 1966 [Non-Patent Document 7] “BERT:Pre-training of Deep Bidirectional Transformers for Language Understanding”;Devlin,Jacob;Chang,Ming-Wei;Lee,Kenton;Toutanova,Kristina(11 October 2018) [Overview of the project]

[0009] According to one aspect of the present invention, a method is provided for processing cell sequencing data to generate a data structure suitable for downstream analysis by clustering based on similarities, such as the ability to react to one or more of the same antigens.

[0010] Clustering may be performed by a two-stage process. In the first stage, the arrays are converted into numerical vectors, and these numerical vectors are clustered into initial clusters. In the second stage, the initial clusters are refined into final clusters, which can be done using the array data itself rather than numerical vectors.

[0011] The dimensionality of sequence data may be reduced while discarding redundant information, so as to preserve properties useful for further analysis. In this case, the useful properties may represent the variability of the immune repertoire. To achieve this, a representative sequence can be assigned to each cluster. The representative sequence can be used as a representation of the set of N sequences within the corresponding cluster, which can be used to simplify any downstream analysis without losing the diversity of the TCR repertoire.

[0012] The output resulting from this method may include a set of clusters, each containing sequences with high sequence similarity and possibly similar functionality. Each cluster can be analyzed individually to accurately identify similar sequences within the repertoire, or the entire set of clusters can be used to analyze the density, sparseness, and other metrics of the TCR repertoire.

[0013] Next, several exemplary embodiments will be described with reference to the accompanying drawings.

Brief Description of the Drawings

[0014] [Figure 1] It is a flowchart showing a method for processing TCR sequence data in an embodiment of the present invention. [Figure 2] An example of a score value in the BLOSUM62 matrix is shown. [Figure 3] It is a diagram showing an example of a computer system capable of executing this method.

Modes for Carrying Out the Invention

[0015] As shown in FIG. 1, the method in one embodiment includes the following steps. 1. Each TCR sequence is converted into a numerical vector, for example, using a natural language processing (NLP) model or by embedding the evolutionary or structural characteristics of the sequence. 2. The TCR sequence vectors are clustered into initial superclusters, for example, using a fast K-means clustering algorithm. 3. Refine the clusters to create final clusters and noise clusters. 4. One or more additional clustering refinement steps. 5. Reduce the cluster dimensions.

[0016] Steps 4 and / or 5 above may be optional depending on the use of the method. Each of these steps will be described in more detail below.

[0017] The TCR sequence may include CDR3 sequence data representing the CDR3 sequences derived from the TCR repertoire. Each item of the CDR3 sequence data may include a string, and examples thereof are shown below: SEQ1. CASSPSGTPYEQYF SEQ2. CASSPSGTTTPYEQYF SEQ3. CASSPSGGTTPYEQYF SEQ4. ASSQASQQQQF SEQ5. ASSSQQQQF SEQ6. QQWQQQSAPA SEQ7. SARAPPYPYNEQF

[0018] The above are a few exemplary arrays, but the method is applicable to processing a very large number of arrays.

[0019] In step 1, each item of the TCR array data is converted into a corresponding numerical vector for efficient processing. This can be done in many different ways. One example is SeqVec based on an NLP (Natural Language Processing) model using ELMo (Embeddings from Language Models), as described in, for example, Non-Patent Document 2. This model moves the sequences into 1024-dimensional vectors that contain their amino acid biophysical information and can be used for downstream data processing. Another NLP model that can be used is a masked language model such as BERT (Bidirectional Encoder Representations from Transformers), as described in, for example, Non-Patent Document 7.

[0020] An alternative approach for step 1 is to use matrix multiplication. Two matrices are selected as value matrices. The first value matrix may be, for example, BLOSUM (BLOcks Substitution Matrix), as disclosed in Non-Patent Document 3. BLOSUM assigns a score for substituting the i-th amino acid with the j-th amino acid. The score value may be, for example, the score value of BLOSUM62 based on a threshold identity of 62% or less, as shown in, for example, Figure 2.

[0021] The second value matrix may be a Miyazawa matrix, for example, as disclosed in Non-Patent Document 4. By multiplying these matrices, output matrices of the same size can be generated for sequences of different lengths. Since both the BLOSUM62 matrix and the Miyazawa matrix contain 21 values ​​per amino acid, the result is a vector of size 441. This vector is used as input for the next step.

[0022] In the example of SEQ1 above, the corresponding numerical vector is a one-dimensional vector with values ​​ranging from -1 to +1, such as [0.8, -0.2...0.1, -0.73].

[0023] In step 2, numerical vectors representing the arrays are clustered together. To efficiently process large amounts of data, clustering is divided into multiple stages. In the first stage, the vectors are clustered into the initial "superclusters," which are later divided into final clusters. The first stage can be performed using a K-means clustering algorithm such as FAISS (https: / / faiss.ai / ), which provides a sufficient approximation for the initial clustering.

[0024] In the above example array, the initial cluster can be defined as follows: SC1:SEQ1-3 SC2:SEQ4-6 SC3:SEQ7

[0025] In step 3, the initial clusters are refined to form the final clusters, for example using HDBSCAN (see Non-Patent Document 5), which is a density-based algorithm that is expected to find some sequence motifs that have a higher density (similar sequences) compared to others, based on the underlying signals in the sample. In this step, clustering is performed on the sequence data, i.e., the sequence data string provided as input in step 1, rather than the corresponding numerical vectors generated in step 1 and clustered in step 2.

[0026] Amino acid sequence clustering can be performed using the Levenshtein distance (see Non-Patent Literature 6), which indicates how many edits are needed for one string to become another. Amino acid sequences with edit distances of 1–2 are expected to maintain very similar functionality and binding affinity to epitopes.

[0027] Therefore, by using the interreceptor distance matrix as input to HDBSCAN, clusters of very similar sequences can be obtained both at the functional level (from the numerical vector in step 2) and for sequence identity (using the Levenshtein distance in step 3).

[0028] A density-based definition can be used for the final cluster. This allows the method to work with clusters of different sizes and sparseness, and downstream, it considers both high-density clusters (many similar receptors) and receptors that are unique within the repertoire.

[0029] If the sequence is not assigned to any cluster in Step 3, it is assigned to the Noise Class (NC), which is as follows: a) Sequences that were incorrectly assigned to the wrong supercluster and therefore did not end up together with similar sequences, or b) A very different arrangement from the rest of the repertoire Includes all sequences that were one of the following.

[0030] In the above example array, the final cluster can be defined as follows: FC1:SEQ1~3 FC2:SEQ4~5 FC3:SEQ6 NC:SEQ7

[0031] Additional refinement steps can be performed after step 3. For example, in step 4, all sequences assigned to the noise class are processed again using steps 2 and 3 to generate new clusters that are added to the final cluster. Sequences that are still in the noise class at this point are defined as clusters of size 1. As an alternative or additional refinement step 4, step 3 (density-based clustering) can be performed again on the final cluster, including any new clusters.

[0032] As an alternative or additional refinement step 4, clusters with high mean pairwise-Levenshtein distances (e.g., exceeding a given threshold) can be dissolved, i.e., member sequences are removed from the cluster and the cluster is deleted.

[0033] One or more refinement steps may be performed iteratively until one or more criteria are met, for example, until the change in the iterative clustering is minimal or no longer changes.

[0034] The output of the method after Step 3 (and optionally Step 4) may be a data file containing a set of N clusters, each containing sequences with high sequence similarity and likely functional similarity. These can be analyzed as further described below, but optionally, the data representing each final cluster can be reduced as shown below.

[0035] In step 5, the dimensionality of each final cluster FC is reduced, for example, by selecting one or more representative sequences from each final cluster. Then, instead of analyzing all sequences from each final cluster, the representative sequences can be used for subsequent analyses.

[0036] In one option, the most expanded sequence of each final cluster is selected as the representative sequence for that cluster. If multiple sequences have the same maximum expansion value, one of them is selected as the representative sequence. For example, the first of several sequences added to the final cluster is selected as the representative sequence.

[0037] Alternatively, the selection of the representative array is based on string distance. A matrix containing each pair of Levenshtein distances is calculated. The sum of the distances for each array is calculated, and the array with the smallest sum is selected as the representative array. If multiple arrays have the same minimum sum, one of the arrays is selected as the representative array. For example, the first of several arrays added to the final cluster is selected as the representative array. Potential applications

[0038] The resulting sequence and the datasets for each cluster have been optionally dimensionally reduced as described above and have many potential applications in research, diagnosis, and / or treatment, examples of which are shown below.

[0039] The dataset can be used for prediction and / or classification in clinical scenarios. For example, clusters may indicate exposure to a specific pathogen, or they may be used to discover novel antigens that could be used in disease-specific drug development.

[0040] The resulting sequence dataset and its respective clusters can optionally be used for other purposes, such as sequence queries, along with the descriptive data and / or reduced dimensions described above. Levenshtein distance can be used to search for specific sequences or similar motifs present in the sample. This allows for rapid sample-to-sample signaling, which may help define public responses to specific antigens.

[0041] Vaccine Development: TCR CDR3 clusters can be used to design and optimize vaccines. By identifying TCR CDR3 clusters associated with protective immune responses against specific pathogens, researchers can develop vaccines that induce targeted immune responses and enhance vaccine efficacy.

[0042] Diagnosis: TCR CDR3 clusters can be used as a diagnostic marker for diseases associated with abnormal immune responses, such as autoimmune diseases or cancer. The presence or absence of specific TCR CDR3 clusters in patient samples can be used to diagnose disease status, monitor disease progression, or evaluate response to immunotherapy.

[0043] Treatment: TCR CDR3 clusters can be used in the development of TCR-based immunotherapies, such as TCR-modified T cell therapy. By incorporating TCR CDR3 clusters, which have high affinity and specificity for tumor antigens or viral antigens, into T cells, genetically modified T cells can be created that can recognize and eliminate cancer cells or infected cells.

[0044] Immune Repertoire Profiling: The TCR CDR3 cluster can be used to profile the diversity and dynamics of the immune repertoire in response to different diseases, infections, or treatments. This can provide valuable information about the clonal diversity, expansion, and contraction of T cell populations, which can aid in understanding immune responses and developing immunotherapies.

[0045] Drug Discovery: TCR CDR3 clusters can be used in drug discovery efforts to develop small molecules or biologics that specifically target TCR-mediated immune responses. By understanding the interactions between TCR CDR3 clusters and their corresponding antigens, researchers can design therapeutics that modulate TCR signaling or function.

[0046] The above are just a few examples of the potential uses of high-quality TCR CDR3 clusters. The field of TCR research is rapidly evolving, and as our understanding of the biology and immunology of TCRs progresses, further applications and discoveries are likely to emerge. Computer system

[0047] Figure 3 shows an exemplary computer system 900 in which the present invention or a part thereof may be implemented as computer-readable code for programming the processing components of the computer system 900. Various embodiments of the present invention will be described in reference to this exemplary computer system 900. For example, each step of the method in Figure 1 can be carried out in system 900. After reading this description, methods for carrying out the present invention using other computer systems and / or computer architectures will become apparent to those skilled in the art.

[0048] The computer system 900 includes one or more processors, such as a processor 904. The processor 904 may be a dedicated processor or a general-purpose processor. The processor 904 is connected to a communication infrastructure 906 (e.g., a bus or network). The computer system 900 also includes a user input interface 903 connected to one or more input devices 905, and a display interface 907 connected to one or more output devices 909, such as a display, which can integrate input and display components. The input devices 905 may include, for example, a connection device such as a mouse or touchpad, a keyboard, or a touchscreen such as a resistive or capacitive touchscreen.

[0049] The computer system 900 also includes main memory 908, preferably random access memory (RAM), and may also include secondary memory 910. The secondary memory 910 may include, for example, a hard disk drive 912, a removable storage drive 914, flash memory, a memory stick, and / or any similar non-volatile memory mechanism. The removable storage drive 914 may include a floppy disk drive, a magnetic tape drive, an optical disk drive, flash memory, etc. The removable storage drive 914 reads from and / or writes to the removable storage unit 918 in a well-known manner. The removable storage unit 918 may include a floppy disk, magnetic tape, optical disk, etc., which is read from and written to by the removable storage drive 914. As will be understood by those skilled in the art, the removable storage unit 918 includes a non-temporary computer-usable storage medium storing computer software and / or data.

[0050] In alternative embodiments, the secondary memory 910 may include other similar means for enabling computer programs or other instructions to be loaded into the computer system 900. Such means may include, for example, a removable storage unit 922 and interface 920. Examples of such means may include a program cartridge and cartridge interface (such as those found in video game devices), a removable memory chip (e.g., an EPROM or PROM) and associated socket, as well as other removable storage units 922 and interfaces 920 that enable the transfer of software and data from the removable storage unit 922 to the computer system 900.

[0051] The computer system 900 also includes a communication interface 924, which may be implemented at the operating system level, for example, to enable the transfer of data between the computer system 900 and external devices, for example, as a signal 928 on a communication channel 926. The communication interface 924 may include a modem, a network interface (such as an Ethernet card), a communication port, a PCMCIA slot and card, and the like.

[0052] Various aspects of the present invention, such as one or more of the method steps described above, may be implemented by software and / or firmware (also called computer programs, instructions, or computer control logic) for programming programmable hardware, or by hardware including dedicated hardwired circuits such as application-specific integrated circuits (ASICs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), or a combination thereof for the computer system 900. Computer programs used to implement the techniques introduced herein may be stored in machine-readable storage media and may be executed by one or more general-purpose or dedicated programmable microprocessors. The terms “computer program media,” “non-temporary computer-readable media,” and “computer-readable media” as introduced herein may generally refer to media such as the removable storage unit 918, the removable storage unit 922, and the hard disks installed in the hard disk drive 912. Computer program media, computer-readable storage media, and computer-readable media may also refer to memory such as the main memory 908 and the secondary memory 910, which may be memory semiconductors (e.g., DRAM). These computer program products are means of providing software to the computer system 900.

[0053] The computer program is stored in main memory 908 and / or secondary memory 910. The computer program may also be received via the communication interface 924. When such a computer program is executed, it enables the computer system 900 to carry out the invention described herein. In particular, when the computer program is executed, it enables the processor 904 to carry out the processes of the embodiments of the invention described above. Thus, such a computer program represents the controller of the computer system 900. If the invention is carried out using software, the software is stored in a computer program product and can be loaded into the computer system 900 using a removable storage drive 914, interface 920, hard drive 912, or communication interface 924. Alternative Embodiments

[0054] The descriptions of various embodiments are presented for illustrative purposes only and are not intended to be exhaustive or limitful to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the embodiments described.

[0055] Aspects of this embodiment may be embodied as a system, method, or computer program product. Aspects of this disclosure may take the form of a fully hardware embodiment, a fully software embodiment (including firmware, resident software, microcode, etc.), or an embodiment combining software and hardware embodiments, all of which may be commonly referred to herein as “engine,” “module,” “system,” or “computer.” Furthermore, any hardware and / or software technology, process, function, component, engine, module, or system described herein may be implemented as a circuit or set of circuits. Furthermore, aspects of this disclosure may take the form of a computer program product embodied on one or more computer-readable media on which computer-readable program code is embodied.

[0056] Aspects of the present disclosure are described above with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It is understood that each block in a flowchart and / or block diagram, and combinations of blocks within a flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a dedicated computer, or other programmable data processing device for manufacturing a machine. When the instructions are executed via the processor of the computer or other programmable data processing device, they enable the implementation of the functions / operations specified in one or more blocks of the flowchart and / or block diagram. Such a processor may be, but is not limited to, a general-purpose processor, a dedicated processor, an application-specific processor, or a field-programmable gate array.

[0057] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of the systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing one or more specified logical functions. Note that in some alternative embodiments, the functions described in the blocks of a flowchart diagram may be performed in an order different from the order shown in the figure. For example, two blocks shown consecutively may actually be executed substantially simultaneously, depending on the associated functionality, or sometimes the two blocks may be executed in reverse order. Note that each block in a flowchart diagram, and any combination of blocks within a flowchart diagram, may be implemented by a dedicated hardware-based system or a combination of dedicated hardware and computer instructions that performs a specified function or operation.

[0058] While the foregoing applies to embodiments of the present disclosure, other and further embodiments of the present disclosure can be devised without departing from its basic scope, the scope of which is determined by the following claims.

Claims

1. A method for processing TCR or BCR cell sequence data, A step of accessing multiple cell sequence data items, each item representing a complementarity-determining region (CDR) of a TCR or BCR cell sequence, The steps include converting the aforementioned cell sequence data items into corresponding numerical vectors, The steps include clustering the aforementioned numerical vector into multiple initial clusters, The steps include refining the initial cluster based on the corresponding cell sequence data items to generate the final cluster, The steps include outputting clustering data representing the final cluster, Methods that include...

2. The method according to claim 1, wherein the cell sequence data items are converted into corresponding numerical vectors by multiplication of a first and a second value matrix.

3. The method according to claim 2, wherein the first value matrix is ​​a BLOSUM matrix.

4. The method according to claim 2 or 3, wherein the second value matrix is ​​a Miyazawa matrix.

5. The method according to any one of the prior claims, wherein the numerical vector is clustered into the plurality of initial clusters using a K-means clustering algorithm.

6. The method according to any one of the prior claims, wherein any of the numerical vectors that cannot be clustered into one of the initial clusters are assigned to a noise class.

7. The method according to claim 6, wherein the step of refining the initial clusters includes assigning one or more cell sequence data items assigned to the noise class to one or more of the refined clusters.

8. The method according to any of the prior claims, wherein the initial clusters are refined using a density-based clustering algorithm.

9. The method according to claim 8, wherein the density-based clustering algorithm is based on the Levenshtein distance between pairs of cell sequence data items.

10. The method according to any of the prior claims, wherein the refinement step includes processing any cell sequence data items remaining in the noise class using the clustering and refinement steps.

11. The method according to any of the prior claims, wherein the refinement step includes iterative clustering.

12. The method according to any of the prior claims, wherein the elaboration step includes dissolving any clusters having an average pairwise distance above a predetermined threshold.

13. The method according to any of the prior claims, wherein the output clustering data includes each selected sequence data item from each of the final clusters.

14. The method according to claim 13, wherein each of the sequence data items is selected as having the maximum extended value among the sequence data items in the corresponding final cluster.

15. The method according to claim 13, wherein each of the sequence data items is selected to have the smallest total pairwise distance among the sequence data items in the corresponding final cluster.

16. A system configured to perform steps of the method described in any of the prior claims.

17. A computer-readable storage medium containing instructions causing the computer to perform the method described in any of the prior claims.

Images