Method for generating DNA libraries to facilitate detection and reporting of low-frequency variants

Abstract

Disclosed is a method for adding adapters to fragmented nucleic acids for next-generation sequencing, comprising providing a numeric code based on variable adapter molecular barcode lengths on both sides of the fragmented nucleic acid, and identifying reads from the same fragment based on both barcodes. The method and product enable amplification of fragmented nucleic acids when the yield of isolated and fragmented nucleic acids is low, and also enable efficient and reliable detection of low-frequency mutations, including subpopulations of cells within a subject.

Description

[Background technology]

[0001] Fields such as cancer treatment, forensic science, paleogenetics, evolution, and toxicology require highly accurate sequencing and detection of low-frequency mutations. Such mutations may be present in less than 1% of cells, including cancer cells. When analyzing cell-free deoxyribonucleic acid (DNA) fragments from plasma or blood samples, the proportion of tumor cell-derived DNA fragments can be as low as 0.01% of the total cell-free DNA. This low-frequency genetic diversity is difficult to evaluate with conventional next-generation sequencing due to high background error rates not only in sequencing itself but also in the amplification of genomic DNA before sequencing. Circulating tumor DNA fragments can be fragmented to an average length of 140–180 bp (base pairs), which may correspond to only a few thousand amplified copies per millimeter of blood. DNA polymerase is 10 -4 ~10 -6 False insertions can be introduced at a frequency of 10 per base. If these false insertions occur early in the generation of the DNA library, such as during first-strand synthesis, they may become indistinguishable from low-frequency mutations. Furthermore, high-throughput sequencing systems, also known as next-generation sequencing (NGS) systems, typically have a frequency of 10 per base. -2 ~10 -3 If errors occur at a certain rate, and the corresponding mutations occur at a similar or lower frequency, it becomes impossible to detect specific true variants.

[0002] For example, single-cell sequencing, single-stranded molecular barcoding, and circle sequencing may involve determining the sequence of DNA derived from a single strand of DNA. During the first amplification, DNA polymerase may propagate errors to daughter molecules. In single-cell sequencing, one strand of a double helicase can be substituted using random primers along with DNA polymerase having helicase activity. However, the combination of random primers and strand substitution can lead to random priming of the newly copied strand, and therefore to the production of the replica. In this process, the initial misincorporation error is propagated to the replica of the replica. Since all genetic information originates from a single cell, it is impossible to distinguish whether a sequencing read represents an error from the original single-stranded synthesis or a genetic variant.

[0003] Circle sequencing (CircSeq) and single-stranded barcoding can also introduce misinsertions during initial synthesis, and these errors can subsequently propagate to daughter molecules and be misscored as mutations. The same misinsertion error after isolation is less likely to occur in the same DNA sequence from other cells or subclonal populations. Therefore, the original error may not necessarily be identified, explained, and / or corrected by post-hoc analysis, instead resulting in errors that appear to be subclonal mutations.

[0004] In "Enhancing the accuracy of next-generation sequencing for detecting rare and subclonal mutations," Salk et al., in Nature Reviews Genetics, Vol.18, pp.269-285, May 2018, outlines three main error correction strategies for better characterizing low-frequency variants using NGS technology: 1) computational strategies based on filtering unreliable data and / or applying predefined statistical models of sequencing error profiles; 2) experimental strategies to reduce errors caused by pre-sequencing DNA library preparation; and 3) molecular consensus sequencing that applies post-hoc detection and correction of errors in the sequencing reads themselves. The latter method utilizes unique tagging of each DNA fragment with molecular barcodes (also known as molecular tags, unique molecular identifiers (UMIs), or single molecular identifiers (SMIs)) before amplification and sequencing, making it possible to group sequencing reads into families of reads associated with specific tags. This facilitates the explicit detection and correction of errors introduced after tagging, as it is unlikely that the exact same error would be systematically repeated across all amplified and sequenced amplicon copies of a uniquely tagged parent DNA fragment. Salk et al. distinguish exogenous molecular barcodes as random or semi-random sequences that are artificially (physically) incorporated into either PCR primers or sequencing adapters, on the one hand, and distinguish them from endogenous molecular barcodes, which can be identified as fragmentation points (also known as shear points) that are naturally (substantially) present at the ends of DNA molecules when preparing DNA libraries using ligation.To date, three major families of molecular consensus sequencing have been developed: 1) single-strand consensus sequencing, such as SafeSeqS, smMIP, and CiqSed, which tag one or both strands of the parent DNA fragment separately (thus having the limitation that it is not possible to use strand information to group amplicon reads generated from complementary strands in downstream consensus error detection and correction steps); 2) double-strand consensus sequencing, such as Ultrasensitive Deep sequencing or CyberSeq, which tag both strands of the parent DNA fragment with the same molecular identifier so that related reads can be grouped into the same consensus sequence after sequencing; and 3) double-strand sequencing, which introduces randomized double tags complementary to both ends of the original double-stranded DNA fragment. These molecular identifier sequences can be encoded within adapters ligated to each end of the double-stranded DNA so that each end of the double-stranded DNA receives a different molecular identifier sequence. If an error is introduced into one of the two strands of DNA by DNA polymerase during the initial strand synthesis or any subsequent synthesis / amplification steps, the other strand provides a basis for comparison against a set of single-stranded consensus sequences, for example. Once all single-stranded consensus sequences are read during sequencing, the molecular identifier sequences at each end of each strand of the original DNA fragment can be matched during alignment.

[0005] To detect post-isolation errors occurring during the synthesis steps following the first synthesis step, each strand may be aligned with its co-sister by associating sequencing reads that share the same start and / or end positions during single-strand consensus sequence alignment using molecular identifier sequences. Any differences in the read sequences may result from misinsertions during the synthesis steps following the first synthesis step. To detect post-isolation errors occurring during the first synthesis step, each strand may be aligned (sequence-compared) with its opposite strand partner (again, using molecular identifier sequences) during double-strand consensus sequence alignment. Differences in read sequences observed by such comparison may result from misinsertions during the first synthesis step. If specific differences are observed in both partner strands of DNA having the same molecular identifier sequences at both ends, these specific differences may result from mutations or polymorphisms present in the DNA extracted from the cells. Low-frequency mutations in a subset of cells may be identified during whole-sequence readout alignment by identifying strands that have substantially similar sequences but different molecular identifier sequences.

[0006] Error-correcting DNA barcodes for high-throughput sequencing, JA Hawkins et al, bioRxiv, 7 May 2018, proposes using more than 10^6 unique error-correcting barcodes by constructing a library of DNA adapters designed according to improvements to information theory codes such as Hamming codes, Reed-Solomon codes, or Levenshtein codes. International Publication 2018 / 144159 proposes using 2 to 24 variable-length nucleotides with a constant 3' overhang to construct a library of DNA adapters with another diversity axis to facilitate the identification of DNA sample fragments. While such methods can facilitate the intrinsic correction of substitution, insertion, and deletion errors to some extent, even when the length of the corrupted barcode is unknown, their specific designs do not fully utilize the error-correcting capabilities of downstream sequencing data processing and variant-calling workflows.

[0007] In "A review of somatic single nucleotide variant calling algorithms for next-generation sequencing data, Computational and Structural Biotechnology Journal 16, pp.15-24, Feb 2018", Xu reviewed 46 publicly available variant callers that could be applicable to single nucleotide variant detection, including 4 variant callers that handle UMI-based sequencing data, probably using duplex and consensus sequencing. As reported by Xu, one limitation of current duplex sequencing protocols is that in actual experiments, only 20% of UMIs can match the other strand due to insufficient ligation efficiency, so variant calling has to handle both single and duplex UMIs. Furthermore, the UMI sequences themselves are prone to PCR errors and may require complementary clustering strategies.

[0008] For example, there is still a need for improved methods to generate DNA libraries that can be linked to the identification of integrated low-frequency variants, in some cases independent of explicit molecular barcoded consensus sequencing error identification / correction, by tracking both strands of a double-stranded DNA sample (such as genomic DNA fragments) to detect very low-frequency mutations and polymorphisms. For example, there is still a need for efficient and reliable methods to detect rare mutations or low-frequency mutations, as well as polymorphisms, in cancer cells, chimeric cells, and other forms of somatic gene polymorphisms. There is also still a need for improved methods to generate DNA libraries that can track both strands of the same DNA molecule and facilitate the identification and reporting of multiple low-frequency variants without requiring explicit consensus sequencing. There is also still a need for improved methods to generate asymmetric fragmented DNA libraries with different sequence characteristics at each end of the DNA fragment to be sequenced or analyzed.

Prior Art Documents

Patent Documents

[0009]

Patent Document 1

Non-Patent Documents

[0010]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0011]

Means for Solving the Problems

[0012] <� A method for generating a library of DNA adapter products from at least two DNA fragments, wherein each DNA adapter product in the library enables the identification of the parental DNA fragment and the analysis of genomic variants after amplification and sequencing. In a reaction mixture, a first adapter is ligated to one end of a first double-stranded DNA fragment having two ends, and a second adapter is ligated to the other end of the first double-stranded DNA fragment having two ends to generate a first DNA-adapter product. Each adapter contains a plurality of double-stranded or partially double-stranded polynucleotides, and each double-stranded or partially double-stranded polynucleotide contains a spacer sequence at the double-stranded end of the adapter. The first adapter spacer sequence (SS1) has a length L1, and the second adapter spacer sequence (SS2) has a length L2. In the same reaction mixture, a third adapter is ligated to one end of a second double-stranded DNA fragment having two ends, and a fourth adapter is ligated to the other end of the second double-stranded fragment having two ends to generate a second DNA-adapter product. Each adapter contains a plurality of double-stranded or partially double-stranded polynucleotides, and each double-stranded or partially double-stranded polynucleotide contains a spacer sequence at the double-stranded end of the adapter. The third adapter spacer sequence (SS3) has a length L3, and the fourth adapter spacer sequence (SS4) has a length L4. Each adapter spacer sequence (SS1, SS2, SS3, SS4) contains a spacer sequence cut from a common fixed predetermined nucleotide sequence (S) of nucleotides of length L S such that the adapter spacer sequences (SS1, SS2, SS3, SS4) differ from each other in length by at least 3 nucleotides and at most L max nucleotides, and L max is L S or more. A method is proposed. The predetermined nucleotide sequence length L S can be between 5 and 20 nucleotides. Each adapter spacer sequence (SS1, SS2, SS3, SS4) contains a cut spacer partial sequence of a fixed length L of at least 3 nucleotides TSA spacer subsequence may be formed by ligating with a constant termination subsequence TS having a constant termination subsequence TS that differs from a certain predetermined nucleotide sequence S by at least 2 in edit distance. The spacer subsequence may be cleaved from left to right starting from a certain nucleotide sequence (S), or cleaved from right to left ending from a certain nucleotide sequence (S). The constant termination subsequence TS may preferably be a triplet nucleotide or a quadruplet nucleotide ending with a T overhang to facilitate ligation to a DNA fragment.

[0013] A method for generating a library of DNA adapter products from at least two DNA fragments to facilitate fragment identification in a genome data analysis workflow of high-throughput sequencing data after amplification and sequencing, comprising the step of generating a pool of DNA adapters, wherein the adapters have a total length of at least 3 nucleotides and a maximum length of L max Unlike other nucleotides, each adapter has a length of L. TS Includes a stationary termination subsequence TS, L TS Three or more nucleotides are linked to the variable spacer subsequence, and the variable spacer subsequence has a length of L S It is cleaved from a common, fixed, predetermined nucleotide sequence (S) having nucleotides, and 5 ≤ L SA step comprising producing a first DNA-adapter product by ligating first and second adapters from a pool of DNA-adapters to the respective ends of a first double-stranded DNA fragment in a reaction mixture having ≤20 nucleotides, wherein each adapter comprises a plurality of double-stranded or partially double-stranded polynucleotides, and each double-stranded or partially double-stranded polynucleotide comprises a spacer sequence at the double-stranded end of the adapter, and as a result, the first DNA-adapter product may be characterized by a numerical code formed by the respective lengths (L1, L2) of the first and second DNA-adapter spacer sequences (SS1, SS2), A method is proposed comprising the steps of producing a second DNA-adapter product by ligating third and fourth adapters from a pool of DNA-adapters to each end of a second double-stranded DNA fragment in the same reaction mixture, wherein each adapter comprises a plurality of double-stranded or partially double-stranded polynucleotides, and each double-stranded or partially double-stranded polynucleotide comprises a spacer sequence at the double-stranded end of the adapter, and as a result, the first DNA-adapter product can be characterized by a numerical code formed by the lengths (L3, L4) of the first and second DNA-adapter spacer sequences (SS3, SS4).

[0014] DNA-adapter products can be amplified to generate PCR replicas, which can then be sequenced to produce raw sequencing reads. Each sequence sequence read R n Regarding this, the genome data analyzer reads from the start of the read to L max =L S +L TS Nucleotides can be trimmed to generate trimmed sequencing reads. The genome data analyzer will then analyze the first L of the sequencing read. max Search for stationary termination subsequences (TS) within nucleotides and perform sequencing reads R n Spacer sequence SS is defined as a function of the number of nucleotides separating the starting point of the stationary termination subsequence TS from the starting point of the spacer sequence TS. Rn Length L n Measure from the start of the lead L nNucleotides can be trimmed to generate trimmed sequencing reads. A genome data analyzer can then analyze the trimmed sequencing reads and, if applicable, the measured length L. n The data can be recorded in a pre-processed sequencing read file, and the trimmed sequencing reads can be aligned to a reference genome, mapping each trimmed read to its start and end positions. The genome data analyzer can use the variable adapter length information measured for each read to facilitate the identification of genomic variants in strands and fragments, respectively, using consensus sequencing or probabilistic sequencing bioinformatics methods. [Brief explanation of the drawing]

[0015] [Figure 1] Figure 1 is a schematic diagram of a genome analysis workflow that includes a tagging step using a ligation adapter to uniquely encode input DNA fragments into DNA-adapter products in a laboratory process (also known as a "wet lab" process), and a pretreatment step for the resulting DNA-adapter product sequencing reads in a bioinformatics workflow (also known as a "dry lab" process) to uniquely identify the DNA fragment source for each read. [Figure 2] Figure 2 is a schematic diagram of an exemplary DNA adapter product for use in DNA library generation. [Figure 3] Figure 3 shows the variety of adapters having variable-length spacer arrays partially cut from a predetermined fixed array, so that they can be used in the proposed method. [Figure 4] Figure 4 shows examples of numerical codes that can be generated by the proposed method for DNA adapter products associated with each DNA fragment. [Figure 5a] Figure 5a) shows an example of a series of spacer arrays SS formed by linking terminal arrays and variable-length cut spacer subarray derivative Si for manufacturing the adapter used in the proposed method. [Figure 5b] Figure 5b) shows an example of a series of spacer arrays SS formed by linking terminal arrays and variable-length cut spacer subarray derivative Si for manufacturing the adapter used in the proposed method. [Figure 6] Figure 6 shows examples of various DNA-adapter PCR replications during the sequencing stage, where the two replications can be tracked back to the same parent DNA product by their unique numerical codes, which can be generated by the proposed method for each DNA-adapter product associated with each DNA fragment. [Figure 7] Figure 7 shows an example of pre-processing raw sequencing reads to identify source DNA fragments and tagging each read accordingly. [Figure 8] Figure 8 shows schematic diagrams of two different possible genome analysis workflow steps for further identifying variants from reads tagged according to the proposed method. [Figure 9] Figure 9 shows the density distribution of each variable-length adapter in the library manufactured according to the proposed method. [Figure 10] Figure 10 shows the ratio of reads assigned to the expected adapter array after sequencing. [Figure 11a] Figure 11a) shows a screenshot of the NGS data viewer for aligned and grouped reads, without considering the proposed adapter numerical code tagging information. [Figure 11b] Figure 11b) shows a screenshot of the NGS data viewer for the same reads aligned and grouped according to the proposed adapter numerical code tagging information to facilitate the identification of heterogeneous SNPs. [Figure 12a] Figure 12a) compares the variant calling results obtained when using the conventional adapter or the proposed variable-length adapter, respectively. [Figure 12b]Figure 12b) compares the variant calling results obtained when using the conventional adapter or the proposed variable-length adapter, respectively. [Figure 12c] Figure 12c) compares the ROC curves for consensus sequencing and stochastic sequencing workflows when using conventional adapters. [Figure 12d] 12d) compares the ROC curves for consensus sequencing and stochastic sequencing workflows when using the proposed variable-length adapter. [Modes for carrying out the invention]

[0016] The details provided herein are illustrative and are intended solely for illustrative purposes of various embodiments, and are presented to provide the most useful and readily understandable explanation of the principles and conceptual aspects of the methods and compositions described herein. In this regard, no attempt has been made to provide more detail than is necessary for a basic understanding, and this explanation will make it clear to those skilled in the art how some forms can actually be embodied.

[0017] The proposed methods and systems will now be described with reference to more detailed embodiments. However, the proposed methods and systems may be embodied in different forms and should not be construed as being limited to the embodiments described herein. Rather, these embodiments are provided to ensure that this disclosure is thorough and complete and fully conveys its scope to those skilled in the art.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art to which the invention pertains. The terms used herein are intended to describe only specific embodiments and are not intended to limit them. As used in the specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural form unless the context clearly indicates otherwise.

[0019] Unless otherwise indicated, the numerical parameters described in the following specification and the attached claims are approximations that may vary depending on the desired characteristics to be obtained, and are therefore modifiable by the term “approximately.” At the very least, and not as an attempt to limit the application of the doctrine of equivalents to the claims, each numerical parameter should be interpreted in light of the number of significant figures and common rounding methods.

[0020] Although numerical ranges and parameters representing a wide range are approximations, the numerical values ​​shown in specific examples are reported as accurately as possible. However, any numerical value inherently contains a certain error that inevitably arises from the standard deviation observed in each test measurement. All numerical ranges given herein include all narrower numerical ranges that fall within such wider numerical ranges, as if all such narrower numerical ranges were explicitly stated herein.

[0021] definition "DNA sample" refers to a nucleic acid sample derived from an organism, such as one that can be extracted from a solid tumor or a fluid. The organism may be a human, animal, plant, fungus, or microorganism. Nucleic acids may be found in limited amounts or low concentrations, such as circulating fetal DNA (cfDNA) or circulating tumor DNA in blood or plasma. "DNA sample" is also used herein to describe RNA samples that have been reverse transcribed and converted to cDNA.

[0022] A "DNA fragment" refers to a short fragment of DNA resulting from the fragmentation of high molecular weight DNA. Fragmentation may occur naturally in the sample organism or may be artificially generated from DNA fragmentation methods applied to the DNA sample, such as mechanical shearing, sonication, enzymatic fragmentation, and other methods. After fragmentation, the DNA fragments can be repaired at the ends to ensure that each molecule has a blunt end. To improve ligation efficiency, adenine may be added to each of the 3' blunt ends of the fragmented DNA to enable ligation of the DNA fragments to an adapter with complementary dT overhangs.

[0023] "DNA products" refer to manipulated fragments of DNA resulting from manipulating, extending, ligating, replicating, amplifying, copying, editing, and / or cutting DNA fragments for application in next-generation sequencing workflows.

[0024] "DNA adapter products" refer to DNA products that result from linking DNA fragments with DNA adapters to adapt them to next-generation sequencing workflows.

[0025] A "DNA library" refers to an assembly of DNA products or DNA adapter products used to adapt DNA fragments for compatibility with next-generation sequencing workflows.

[0026] A "pool" refers to multiple DNA samples (e.g., 48, 96, or more) from the same or different organisms that can be multiplexed into a single high-throughput sequencing analysis. Each sample may be identified within the pool by a unique sample barcode.

[0027] A “nucleotide sequence” or “polynucleotide sequence” refers to any polymer or oligomer of nucleotides, such as cytosine (represented by the letter C in a sequence), thymine (represented by the letter T in a sequence), adenine (represented by the letter A in a sequence), guanine (represented by the letter G in a sequence), and uracil (represented by the letter U in a sequence). A “nucleotide sequence” or “polynucleotide sequence” may be DNA or RNA, or a combination thereof. A “nucleotide sequence” or “polynucleotide sequence” may be found permanently or transiently in single-stranded or double-stranded form. Unless otherwise indicated, nucleic acid sequences are written from left to right in the 5' to 3' direction.

[0028] A "random sequence" or "partially random sequence" refers to a sequence of nucleotides selected at least partially randomly from all possible combinations of nucleotides for a given sequence length. The selection of a random sequence can be manual or automatic.

[0029] A “fixed sequence” or “predetermined sequence” refers to a fully specified, non-random, fixed nucleotide sequence that is specifically selected from all possible combinations of nucleotides for a given sequence length. The selection of the non-random sequence may be manual or automatic. The selection of the non-random sequence may be based on specific criteria specific to the sequencing application and / or sequencing technology, for example, to enhance the error robustness of the amplification and sequencing processes.

[0030] A "primer sequence" refers to a nucleotide sequence of at least 20 nucleotides in length, which includes a region complementary to the target DNA to be extended or amplified, in whole or in part.

[0031] The "edit distance" between two nucleotide sequences refers to the minimum number of nucleotide substitutions, insertions, or deletions that must be applied for one sequence to become identical to the other.

[0032] "Luncapping" refers to the joining of separate double-stranded DNA sequences. The latter DNA molecules may have blunt ends or may have overhangs that are adapted to facilitate ligation. Linchapping can be produced by various methods, such as using ligase enzymes, performing chemical ligation, and other methods.

[0033] "Amplification" refers to a polynucleotide amplification reaction that generates multiple polynucleotide sequences replicated from one or more parent sequences. Amplification can be achieved by various methods, such as polymerase chain reaction (PCR), linear polymerase chain reaction, nucleic acid sequence-based amplification, rolling circle amplification, and other methods.

[0034] "Sequencing" refers to reading the sequence of nucleotides as a string. High-throughput sequencing (HTS) or next-generation sequencing (NGS) refers to the real-time sequencing of multiple sequences, typically 50 to several thousand base pairs long, in parallel. Exemplary NGS technologies include those from Illumina, Ion Torrent Systems, Oxford Nanopore Technologies, Complete Genomics, and Pacific Biosciences. Furthermore, to facilitate the sequencing and amplification processes, depending on the actual technology, NGS sequencing may require sample pretreatment using sequencing adapters or primers. This allows, for example, in the case of synthesis sequencing, multiple examples of a single parent molecule to be sequenced by PCR amplification before delivery to the flow cell.

[0035] An "adapter" refers to a short, double-stranded or partially double-stranded DNA molecule of approximately 10 to 100 nucleotides (base pairs) designed to be ligated to a DNA fragment. Adapters may have a blunt end, an overhang as a 3' or 5' overhang, or a combination thereof. For example, to improve ligation efficiency, an adenine may be added to each of the 3' blunt ends of the fragmented DNA before adapter ligation, and the adapter may have a thymidine overhang at its 3' end to base-pair with the adenine added to the 3' end of the fragmented DNA. Adapters may have a phosphorothioate bond before the terminal thymidine at the 3' end to prevent exonucleases from trimming the thymidine, thus forming a blunt end when the end of the adapter being ligated is double-stranded.

[0036] A "partially double-stranded adapter" refers to an adapter that contains both a double-stranded region and a single-stranded region. The double-stranded region of the adapter contains a linking domain, while the single-stranded region contains primer sequences used for subsequent library amplification, barcoding, and / or sequencing. The single-stranded region may consist of two single-stranded arms, a 5' arm and a 3' arm, as in the case of a so-called Y-shaped adapter, or the single-stranded region of a partially double-stranded adapter may form a hairpin or loop, as in the case of a so-called U-shaped adapter. Therefore, the term partially double-stranded adapter refers to both Y-shaped adapters and U-shaped adapters or combinations thereof.

[0037] A "PCR replica" refers to a replica generated by PCR amplification from a single-stranded DNA molecule belonging to a DNA-adapter product derived from the original DNA fragment.

[0038] A "molecular tag," "molecular barcode," "molecular code," or "molecular identifier" refers to a molecular sequence, such as a nucleic acid sequence, that is fully and uniquely specified by its nucleotide string.

[0039] A “numerical code,” “non-molecular code,” or “non-molecular identifier” refers to a numerical measurement of one or more intrinsic properties of a molecular sequence that are not the molecular sequence itself. Examples of properties of nucleic acid molecular sequences include length, size, molecular weight, molar concentration, polarity, elasticity, stiffness, electrical conductivity, fluorescence, reflectivity to a particular excitation wave, or, more generally, any physical, chemical, or biological properties that can be experimentally measured for the molecular sequence and / or a portion of the molecular sequence.

[0040] A "variable-length code (VLC)" refers to the variable length of a nucleic acid sequence that can be measured as the number of nucleotides, monomers, polymers, homopolymers, heteropolymers, or combinations thereof.

[0041] "Read trimming" or "read preprocessing" in a bioinformatics workflow refers to the process of removing sets of nucleotides at the starting points of a read sequence string, such as nucleotides corresponding to adapter sequences, from the sequencing reads to extract the actual DNA fragment sequences to be analyzed.

[0042] "Aligning," "alignment," or "aligner" in a bioinformatics workflow refers to mapping and aligning base by base, where pre-processed sequencing reads are read from a reference genome sequence, depending on the application. For example, in a targeted enrichment application where sequencing reads are expected to map to a specific target genomic region according to a hybrid capture probe used in the experimental amplification step, alignment may be specifically searched for a corresponding sequence defined by genomic coordinates such as chromosome number, start position, and end position within the reference genome.

[0043] "Variant calling," or "variant calling," refers to the process of identifying actual variants within aligned reads in a bioinformatics workflow. Variants may include single nucleotide substitutions (SNPs), insertions or deletions (INDELs), copy number variants (CNVs), as well as major rearrangements, substitutions, duplications, and translocations. Preferably, variant calling is robust enough to discriminate actual variants from amplification and sequencing noise artifacts.

[0044] "Consensus sequencing" in a bioinformatics workflow refers to the process of grouping sequencing reads into families of reads generated from the same double-stranded DNA fragment and / or the same DNA fragment strand, comparing them to detect errors in the amplification and / or sequencing process, and correcting those errors to generate a unique deterministic consensus sequence for the double-stranded DNA fragment or DNA fragment strand. Variant calling is then performed by processing the resulting consensus sequence, rather than the entire read.

[0045] "Probabilistic sequencing" in bioinformatics workflows refers to directly performing variant calling on the data by processing the entire set of reads from different families to group sequencing reads into families of reads generated from the same double-stranded DNA fragment and / or the same DNA fragment strand, and then comparing the data to a probabilistic model to calculate the probability of the data supporting all possible genotypes at each genomic location being analyzed.

[0046] Workflow Next, with reference to Figure 1, a typical workflow for identifying low-frequency DNA variants will be described in more detail. As will be apparent to those skilled in the field of DNA analysis, such a workflow includes preliminary experimental steps performed in a laboratory (also known as a “wet lab”) to generate DNA analysis data, such as raw sequencing reads in a next-generation sequencing workflow, as well as subsequent data processing steps performed on the DNA analysis data to further identify information of interest to the end user, such as detailed identification of DNA variants and associated annotations, using a bioinformatics system (also known as a “dry lab”). Depending on the actual application, laboratory setup, and bioinformatics platform, various embodiments of the DNA analysis workflow are possible. Figure 1 illustrates an example workflow that includes a wet lab step in which a DNA sample is first fragmented using fragmentation protocol 50 (optional) to generate DNA fragments. The DNA ends of these DNA fragments are then repaired and modified to fit the adapter to be used. The adapter, such as one described in more detail through this enclosure, can then be ligated to the DNA fragments in the reaction mixture by ligation 100 according to some of the proposed methods, thereby generating a library of DNA-adapter products. The DNA library then undergoes amplification 110 and sequencing 120. In the next-generation sequencing workflow, the resulting DNA analysis data can be generated as a data file of raw sequencing reads in FASTQ format. The workflow may then further include a drylab genome data analyzer system 150, which takes the raw sequencing reads into input to a pool of DNA samples prepared using a ligation adapter according to the proposed method and applies a series of data processing steps to identify genome variants, for example, as a genome variant report for the end user. An exemplary genome data analyzer system 150 is the Sophia Data-Driven Healthcare Platform (Sophia DDM), which was already in use in over 1,000 hospitals worldwide in 2019, but other systems can be used in a similar manner.Various detailed possible embodiments of the data processing steps that can be applied by the genome data analyzer system 150 are described, for example, in International PCT Patent Application Publication WO2017 / 220508, but other embodiments are also possible.

[0047] In a preferred embodiment, the genome data analyzer system 150 may first apply one or more pre-processing steps 151 to generate pre-processed reads from raw sequencing read inputs. The pre-processing steps may include, for example, adapter trimming and read sorting for analyzing and grouping families of reads generated from similar DNA fragments according to proposed adapter concatenation and numerical coding methods, as further described herein. In possible embodiments, both raw and pre-processed reads may be stored in a FASTQ file format, but other embodiments are also possible.

[0048] The genome data analyzer system 150 can further apply the sequence alignment 152 to the pre-processed reads to generate read alignment data. In one embodiment, the read alignment data may be generated in, for example, a BAM or SAM file format, but other embodiments are also possible.

[0049] The genome data analyzer system 150 can further apply variant calling 153 to read alignment data to generate variant calling data. In one embodiment, the variant calling data may be generated in, for example, a VCF file format, but other embodiments are also possible.

[0050] The genome data analyzer system 150 can further apply variant annotation 154 to the read alignment data to create a genome variant report for each DNA sample. In one embodiment, the genome variant report can be visualized by the end user on a graphical user interface. In another possible embodiment, the genome variant report can be further created as a text file for data processing. Other embodiments are also possible.

[0051] fragmentation In some embodiments, the methods described herein involve the use of sequenced genomic and mitochondrial DNA, and the determination of information such as the location and coding of epigenetic information, including genes, promoters, exons, introns, and potential CpG islands, as well as information such as methylation, potentially combined with bisulfide (disulfide) conversion. Genomic DNA may be chromosomal DNA or circular DNA. Alternatively, mRNA may be reverse transcribed into complementary DNA or cDNA, the cDNA may be fragmented or may be short enough to be sequenced without fragmentation. Fragmented or unfragmented complementary cDNA may be single-stranded and then double-stranded by annealing random primers and / or other primers to extend the primers to be complementary to the cDNA, thus forming double-stranded cDNA. In some embodiments, double-stranded cDNA and mitochondrial DNA and / or genomic DNA must be fragmented before sequencing 120 (50). Fragmentation 50 can be achieved by several means, including but not limited to sonication, mechanical shearing, and partial digestion by restriction enzyme digestion, for example. As a result of fragmentation, the fragmented DNA may be 50 to 10,000 base pairs long, preferably 200 to 800 base pairs long, more preferably 300 to 500 base pairs long, and even more preferably 400 base pairs long. Whether the DNA fragments originate from cDNA, genomic DNA, or mitochondrial DNA, they can be resized by means of, for example, agarose gel electrophoresis; gel chromatography; equilibrium density gradient centrifugation including sucrose gradient centrifugation, Percoll gradient centrifugation, and cesium chloride centrifugation; and other means.

[0052] Adapter connection / insertion In the case of reverse transcription following fragmentation and end repair 50, and subsequent formation of genomic DNA, chromosomal DNA, or double-stranded DNA, the adapter may be ligated or joined to each of the ends of the fragmented double-stranded DNA (100).

[0053] Figure 2 shows an embodiment in which two adapters 200, 250 are linked to each end of a DNA fragment 220 (100). Each adapter 200, 250 shown in the exemplary embodiment shown in Figure 2 may contain a partial double-stranded DNA molecule having a single nucleotide (T) 3' overhang at the end to be annealed to the double-stranded fragmented DNA. Each adapter 200, 250 contains double-stranded segments 210, 260 at one end that constitute a spacer sequence (SS) that separates the adapters 200, 250 from the nucleotide sequence of the DNA fragment 220 in subsequent high-throughput sequencing reads (read 1, read 2). In possible embodiments as illustrated in Figure 2, the latter spacer sequence end may contain a single-stranded nucleotide T3' overhang, but other embodiments are also possible as will be apparent to those skilled in the art, for example, to facilitate the linkage 100 of the adapters 200, 250 to the target double-stranded DNA molecule 220 (e.g., genomic DNA or "gDNA"), they may be blunt ends or substituted with another 3' or 5' overhang.

[0054] The adapter contains a double-stranded sequence at the end annealed to the double-stranded DNA. In this regard, one of the two strands of the adapter's double-stranded sequence is annealed to the 3' end of the fragmented double-stranded DNA, and the other of the two strands of the adapter's double-stranded sequence is annealed to the 5' end of the fragmented double-stranded DNA.

[0055] The ends of the double-stranded sequence of the adapter ligated to the fragmented double-stranded DNA are not limited and may include blunt ends, 3' overhangs, and 5' overhangs. In this regard, the 5' end of the ligated adapter may be terminated with a 5'-phosphate or a 5'-OH. If the 5'-OH is at the adapter end ligated to the target nucleic acid, it may be necessary to use a polynucleotide kinase to complete the backbone and bind the 5'-OH of the adapter to the 3'-OH of the fragmented DNA. In some embodiments, a single nucleotide overhang that can be ligated by a T-4 ligase from a T-4 bacteriophage is preferred. Thus, in some embodiments, an adenine may be added to each of the 3' blunt ends of the fragmented DNA before adapter ligation, and the adapter may have a thymidine overhang at its 3' end to base pair with the adenine added to the 3' end of the fragmented DNA. In some embodiments, an adenine may be added to each of the 3' blunt ends of the fragmented DNA before adapter ligation, and the adapter may have a phosphorothioate bond before the terminal thymidine at the 3' end to base pair with the adenine added to the 3' end of the fragmented DNA. The phosphorothioate bond before the terminal thymidine prevents the exonuclease from trimming the thymidine, resulting in the formation of a blunt end when the ends of the ligated adapter are double-stranded.

[0056] Adapter with variable-length spacer array In a preferred embodiment, as shown in Figure 2, each adapter 200, 250 includes spacer sequences 210, 260 that terminate their double-stranded ends to be attached to the DNA fragment 220. In one embodiment, some or all of the spacer sequences 210, 260 have a length L S A nucleotide is cleaved from a predetermined, fixed nucleotide sequence S, resulting in various variable-length cleavage type spacer subsequences S. i S j It can form.

[0057] In one embodiment, as shown in Figure 2, to facilitate downstream read trimming preprocessing 151 from raw sequencing reads by the genome data analyzer 150, up to L S The length of each nucleotide L Si , L Sj Each of the cut-type spacer partial arrangements S i S j Afterwards, for example, each sectional variable-length subarrangement S i S j By ligating it with the TS termination subsequence, a length of at least 3 nucleotides L is obtained. TS The steady-state termination subarray TS continues, potentially forming variable-length spacer arrays 210 and 260.

[0058] Preferably, a predetermined constant nucleotide sequence S and a constant termination subsequence TS are selected such that the constant termination subsequence TS differs from a reminder of sequence S by at least 2 in edit distance. Thus, as shown in Figure 2, each adapter spacer sequence SS 210, 260 may be terminated by the same constant termination subsequence TS of at least 3 nucleotides, the termination subsequence TS being a reminder of sequence S (and therefore its derived cleavage-type spacer subsequence S). i S j (from any of the above) and with an edit distance that differs by at least 2.

[0059] As shown in Figure 3, multiple adapters may be used, specifically, a partial arrangement S of their cut-type spacers. i S j The lengths of the spacer sequences vary from one another. Therefore, the total length obtained for spacer sequences 210, 260 after ligation with a certain termination subsequence TS can be, for example, 3 nucleotides (similar to the minimum size for termination subsequences that can be used as “triplet stop codes” to facilitate downstream read trimming pretreatment 151), 10(7+3) nucleotides, 5(2+3) nucleotides, 7(4+3) nucleotides, 4(1+3) nucleotides, etc. More generally, the variable length of the spacer sequence is at least L TS =3 nucleotides, maximum Lmax =L S +L TS It can be a nucleotide. Similarly, for quadruplet termination subsequences TS, the variable length of the spacer sequence is at least L TS =4 nucleotides, maximum L max =L S +L TS It could be a nucleotide, for example.

[0060] Generally, the maximum length L of a given polynucleotide sequence S is... S The induced spacer sequences 210, 260 may be selected to allow for sufficiently different variable cleavage lengths to provide the diversity of combinations necessary to distinguish PCR replication from similar DNA fragments, i.e., fragments that share the same reference mapping position once ligated with a pair of adapters from a plurality of adapters having different cleavage lengths, while ensuring that the induced spacer sequences 210, 260 do not take segments that are too long relative to the total sequencing read length (which can be as low as 150 base pairs in some high-throughput sequencing workflows), from a bioinformatics workflow perspective. In possible embodiments, L S The nucleotide sequence (TS,S) can be selected as 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 nucleotides, although other embodiments are possible. When preparing a pool of samples for high-throughput sequencing, in a possible embodiment, the same constant polynucleotide sequence S can be used to prepare ligation adapters for all samples in the pool of samples that will be multiplexed together in the NGS workflow. In another embodiment, different constant polynucleotide sequences can be defined and used to prepare ligation adapters for different samples in the same pool. In the latter embodiment, the multiple adapters generated for each sample in the pool of samples may differ by either a predefined termination subsequence (TS) or a predefined nucleotide sequence (S) used to cleave a variable spacer subsequence. Thus, the selection of (TS,S) is common and constant for all fragments of the sample, but different for each sample in the same pool.

[0061] Figure 4 shows the ligations and corresponding numerical codes obtained for three exemplary DNA fragments 421, 422, and 423. The first DNA fragment 421 is ligated to a first adapter 401 containing a spacer sequence 411 (SS1) with a full-length L1 of 3 nucleotides at one end, and to a second adapter 451 containing a spacer sequence 461 (SS2) with a full-length L2 of 10 nucleotides at the other end. The second DNA fragment 422 is ligated to a third adapter 402 containing a spacer sequence 412 (SS3) with a full-length L3 of 5 nucleotides at one end, and to a fourth adapter 452 containing the same spacer sequence 462 (SS4 - SS4 = SS3 in this specific example) with a full-length L4 = L3 = 5 nucleotides at the other end. The third DNA fragment 423 is linked to a fifth adapter 403 containing a spacer sequence 413 (SS5) with a full-length L5 of 7 nucleotides at one end, and to a sixth adapter 453 containing a spacer sequence 463 (SS6) with a full-length L6 of 4 nucleotides at the other end. Thus, the first DNA-adapter product generated from DNA fragment 421 may associate with the numerical code {3,10} (or {10,3} depending on the read direction) corresponding to the respective lengths of the spacer sequences from the adapters at both ends. Thus, the second DNA-adapter product generated from DNA fragment 422 may associate with the numerical code {5,5} (in any read direction) corresponding to the respective lengths of the spacer sequences from the adapters at both ends. Thus, the third DNA-adapter product generated from DNA fragment 423 may associate with the numerical code {7,4} (or {4,7} depending on the read direction) corresponding to the respective lengths of the spacer sequences from the adapters at both ends. Therefore, it is possible to trace back each derivative DNA product from its parent DNA adapter product by identifying the first, second, and third DNA adapter products that have the same mapping position in the DNA library, identifying the spacer sequences at both ends of the derivative DNA product, measuring the length of each of them, and identifying the numerical code inherited from the parent DNA adapter product.

[0062] Figure 5a) shows a first example of 10 possible spacer arrangements identified as v9, v8, v7, v6, v5, v4, v3, v2, v1, and v0 in Figure 5a). Each spacer arrangement has a length L S It can be formed by a left-to-right cut from the beginning of the stationary sequence S=CCACAACAC of length L = 9, and can further be linked with a termination subsequence (TS) triplet T, G, T (which itself ends with a T overhang to facilitate linking). Figure 5b) shows a sequence of length L S Here is another second example of 10 possible spacer sequences that can be cut from right to left from the end of a constant sequence S=CCACAACAC with length L = 9 and further connected to a terminal subsequence (TS) triplet T, G, T (which itself ends with a T overhang to facilitate connection). In both examples in Figures 5a) and 5b), the constant sequence S=CCACAACAC has length L S =9, and the possible derived cleavage subsequences have subsequence lengths of 9, 8, 7, 6, 5, 4, 3, 2, 1, and 0 nucleotides, respectively. When a triplet of T, G, T nucleotides follows, this corresponds to the triplet code TGT in the resulting sequencing read, and the resulting spacer sequence lengths are 12, 11, 10, 9, 8, 7, 6, 5, 4, and 3 nucleotides, respectively.

[0063] In possible embodiments, the full spacer array length of the truncated positive terminal subarray (absolute length, e.g., a number in the range of 3 to 12) may be used to form a numeric code. In another embodiment, the sole length of the truncated subarray portion of the spacer array, and therefore excluding the constant length of the terminal subarray, may be used to form a numeric code (relative length, e.g., a number in the range of 0 to 7).

[0064] Amplification and sequencing Once a DNA product is produced using adapter linkage, the DNA product can be amplified by a polynucleotide amplification reaction to produce multiple polynucleotide sequences replicated from one or more parent sequences. As will be apparent to those skilled in the art in the field of next-generation sequencing, amplification can be carried out by various methods, such as polymerase chain reaction (PCR), linear polymerase chain reaction, nucleic acid sequence-based amplification, rolling circle amplification, and other methods. In some embodiments, after library amplification, the DNA-adapter product can then be sequenced using any technology known in the art, including but not limited to Illumina sequencing technology, Ion Torrent sequencing technology, 454 Life Sciences sequencing technology, ABI SOliD sequencing technology, Pacific Biosciences sequencing technology, or Oxford nanopore sequencing technology. For example, in the case of the Illumina sequencing platform, the sequencer primer ring sequences present at both ends of the library product have functional properties that allow them to anneal to or bind to flow cell oligomers or flow cell sequences. As will be apparent to those skilled in the field of next-generation sequencing, a bridge amplification step 110 may then be performed, in which fragmented DNA comprising an adapter sequence (including a spacer sequence), a first primer ring sequence, and a second primer ring sequence is annealed to either a first or / or second immobilized sequence. The fragmented DNA comprising the adapter sequence, the first primer sequence, and the second primer sequence is then used as a template to extend the 3'-OH groups of the first and / or second immobilized sequence, thereby transferring the genetic information in the fragmented DNA comprising the adapter sequence (including the proposed spacer sequence), the first primer sequence, and the second primer sequence to the first or second immobilized sequence, thereby binding to a solid support. The fragmented DNA comprising the adapter sequence (including the proposed spacer sequence), the first primer sequence, and the second primer sequence is then denatured or deannealed and removed.Next, the bound and fragmented DNA is annealed to the immobilized sequence at its free ends and undergoes several cycles of bridge amplification.

[0065] At this point, the clustering process is complete, and the flow cell is configured to allow sequencing by synthesis by cleaving the free immobilized sequence and thus re-annealing it to the free immobilized sequence. After priming, each nucleotide can be incorporated into the newly synthesized DNA strand based on the template strand annealed to the solid support during clustering. Each nucleotide incorporated into the newly synthesized strand is associated with a different fluorophore, each fluorophore emitting light of different wavelengths if the newly incorporated nucleotide can be base-paired with the new strand of DNA and / or its complementary counterpart (A to T, G to C) during extension.

[0066] In exonuclease-based nanopore sequencing, nucleic acids can be digested, and the resulting free nucleotides are identified by their effect on the potential across the lipid membrane. Single-stranded nucleic acid strands can also be forced to pass through nanopores driven by the difference in potential or assisted by enzymes such as helicases or polymerases. The movement of nucleic acid strands through the nanopores can result in changes in potential that enable the identification of nucleic acid sequences.

[0067] Next, the index sequence can be used to identify a sample of the sequence. After read pretreatment 151 and read alignment 152, the endogenous information and / or mapping location of the DNA fragment, the exogenous mapping location of the DNA fragment, or a combination thereof can be used to identify PCR replication and distinguish true mutants from misinsertions that occur after DNA fragmentation.

[0068] In some embodiments, during the polymerization or extension of a new strand of DNA from a template strand, DNA polymerase may misplace a base that does not base-pair with the opposite nucleotide of the other strand of DNA, which is called a misplaced base pair or misinsertion. In this regard, even if one or more misplaced base pairs can occur, the newly synthesized DNA strand can be considered complementary to the template strand. In some embodiments, it is intended that this misplaced base pair error by DNA polymerase may occur in the daughter strand of DNA, and that by tracking all copies belonging to the same PCR double-strand group as this daughter strand, it may be possible to distinguish these misplaced base pairs from genetic polymorphisms (e.g., mutations) found in genomic DNA extracted from cells.

[0069] Read preprocessing After amplification 110, each DNA-adapter product is replicated in multiple PCR replications. Thus, as shown in Figure 6, two PCR replicas 601 and 602 generated from the same DNA adapter product, i.e., from the same DNA fragment, have the same start and end coordinates and the same spacer sequence at their ends, which is observed in the raw sequencing reads obtained after sequencing 120. Therefore, by measuring the length of each of their spacer sequences (in the example in Figure 6, numerical code = {9,7}), it is possible to group them together in the downstream genome analysis workflow.

[0070] As those skilled in the art of low-frequency DNA analysis will understand, other DNA-adapter products, i.e., PCR replicas generated from different DNA fragments, are unlikely to have the same spacer sequence length. However, this is provided that 1) the number of possible combinations of different adapters is sufficiently large compared to the number of potentially colliding DNA fragments to distinguish them from reads that have the same start and end positions after alignment 152, and 2) errors in PCR amplification and sequencing, including possible nucleotide insertions or deletions in the spacer sequence, can be detected by using a constant sequence S as the basis for the cleaved spacer sub-sequences recovered in the read.

[0071] As shown in Figure 2, in the case of paired-end sequencing technology, after sequencing, two different read directions, READ1 and READ2, may each generate different spacer sequences with a common termination sequence TS in the FASTQ file. These spacer sequences may have different lengths for each DNA adapter product, and therefore can be statistically distinguished from one another. In the alignment step 152, the start and end positions of the DNA fragment sequences 220 to be analyzed thus move apart among the majority of reads generated from different DNA adapter products, thereby creating further intrinsic diversity.

[0072] For example, referring to Figure 4, for the first DNA fragment 421, the first spacer sequence 411 constitutes the first three nucleotides in the raw sequencing read of the PCR replica read in the 3' to 5' direction, while the second spacer sequence 461 constitutes the first ten nucleotides in the raw sequencing read of the same PCR replica read in the reverse direction from 5' to 3'. For the second DNA fragment 422, the third spacer sequence 412 constitutes the first five nucleotides in the raw sequencing read of the PCR replica read in the 3' to 5' direction, and the fourth spacer sequence 462 constitutes the first five nucleotides in the raw sequencing read of the same PCR replica read in the reverse direction from 5' to 3'. In the case of the third DNA fragment 423, the fifth spacer sequence 413 constitutes the first seven nucleotides of the raw sequencing read of the PCR replica read in the 3' to 5' direction, and the sixth spacer sequence 463 constitutes the first four nucleotides of the raw sequencing read of the same PCR replica read in the reverse direction from 5' to 3'. Thus, it is possible to uniquely associate a numerical code with each DNA fragment, such as the combination of spacer sequence length values ​​for the first and second ends of the first DNA fragment 421 {L1,L2}={3,10}; the combination of spacer sequence length values ​​for the first and second ends of the second DNA fragment 422 {L3,L4}={5,5}; and the combination of spacer sequence length values ​​for the first and second ends of the third DNA fragment 423 {L5,L6}={7,4}. Therefore, as shown in Figure 4, it is possible to group PCR replica raw sequencing reads based on the variable length of spacer sequences, which can be searched for from their start sequences of nucleotides in the raw sequencing reads sequenced from PCR replicas generated from DNA-adapter products produced by the proposed method.

[0073] Figure 7 shows examples of start sequences for three different reads that may be generated from sequencing a DNA-adapter product constructed according to the exemplary sequence in Figure 5a). Each spacer sequence ends in a termination sequence triplet TGT, as in the example in Figure 5a), so the genome data analyzer 150 can search for this triplet as part of the read preprocessing step 151.

[0074] In a first possible embodiment (not shown), the read preprocessing 151 first processes the read start sequence L max It consists of trimming by nucleotides, L max The length L of a certain sequence S from which a subsequence is cut is... S And the length L of the connected constant termination subarray TS TS This is the sum of the above. After trimming the reads in the FASTQ file, the remaining sequence of each read can be stored in the pre-processed FASTQ file.

[0075] As will be apparent to those skilled in the art of sequencing, the use of variable-length adapters results in the resulting pre-treated reads shifting relative to each other at different start and end positions after subsequent alignment 152, thereby effectively separating the alignment results obtained from different DNA adapter products statistically. However, the latter "intrinsic" length identification may not be statistically sufficient to identify the DNA fragments to be analyzed, depending on the requirements of the actual application. Furthermore, it has the disadvantage of losing several nucleotides at the start of the fragment to trim even reads carrying shorter cut spacer subsequences to the longest possible adapter length. Therefore, in another embodiment, a terminal subsequence TS may be searched for at the start of each read sequence. Once found, the length of the spacer sequence string can be measured, for example, as the distance between the start of the read and the start of the termination subsequence TS (relative spacer sequence SS length). Alternatively, it can be measured as the distance between the start of the read and the end of the terminal subsequence TS (absolute spacer sequence SS length). Therefore, each read may be assigned a different spacer sequence length measurement as part of the read pretreatment step 151. In the example in Figure 7, the first read has a spacer sequence SS1=CCACAACACTGT with an absolute length L1=12 at the starting point. The second read has a spacer sequence SS2=ACAACACTGT with an absolute length L2=10, and the third read has a spacer sequence SS3=CTGT with an absolute length L3=4. Thus, the measured length values ​​may be recorded in the pretreated FASTQ file to provide additional numerical information that allows tracing back to the DNA-adapter product origin of the read in the downstream alignment step 152. Therefore, in accordance with the actual requirements of this application, the remainder of the read sequences entered into alignment may generally be assigned the longest possible spacer sequence length L to provide further "intrinsic" length identification to the alignment step. max It can be trimmed to (at the expense of losing a few nucleotides at the start of the fragment sequence itself), or instead, each sequencing read R can be pre-treated by pre-treatment 151. nThe actual spacer array SS length L is measured for n It can be trimmed individually (up to the end of the terminal subsequence TS).

[0076] Lead mapping and alignment The resulting pre-processed reads can then be aligned to a reference genome (152). The set of reads generated from PCR replications of different original DNA fragments can then be identified in the data record (typically stored as a BAM or SAM file format) based on one or more of the following features available in the data record:

[0077] 1) A numerical code obtained by combining the adapter spacer array lengths measured in the lead; 2) The mapping position (i.e., the starting point) of the DNA fragment relative to the reference genome.

[0078] In paired-end sequencing, paired-end read orientation information (i.e., F1R2 or F2R1) can be used to identify paired-end reads generated from the original positive or negative strands. For each pair of paired-end reads (i.e., R1 and R2), their possible different adapter lengths can be recovered, and these numbers can be used to form a numerical code (consisting of pairs of integer values) that is stored as a tag in a read alignment file such as a BAM format file. In the first step, paired-end reads that are aligned to the same start and end positions (with respect to the reference genome sequence read direction) and have the same pair of measured adapter lengths (L1,L2) or (L2,L1) can be grouped as sequencing reads that are likely generated from the two strands of the same original double-stranded DNA fragment. Each group can then be further subdivided into two subgroups according to their original strands, and the actual pair of measured adapter lengths (L1,L2) is {L n(forward) ,L m(reverse)} In the case of paired-end reads with F2R1 orientation, {L n(reverse),L m(forward) It is given by}.

[0079] The obtained information can be recorded in a raw fragment-tagged read alignment file, such as a BAM or SAM format file. Using this file, it is possible to cluster groups of paired-end reads generated from the same fragment ligatures in the alignment, and as a result, downstream genome analysis steps such as variant calling can be performed by utilizing the information provided by PCR replicas generated from the two strands of the original DNA fragment.

[0080] Variant calling Next, the obtained aligned reads can be analyzed (153) to identify variants relative to the reference genome, such as SNVs, indels, or structural variants (copy number variations, duplications, translocations, etc.). Different approaches can be applied by the genome data analyzer 150, as shown in Figure 8, and also as outlined by Xu, for example, in A review of somatic single nucleotide variant calling algorithms for next-generation sequencing data, Computational and Structural Biotechnology Journal 16, pp.15-24, Feb 2018. Figure 8a) shows a consensus sequencing approach in which a single polynucleotide sequence is folded from each group of sequence reads sharing the same alignment position and numerical code tag in the aligned BAM file, according to the proposed method. As shown in the circle in Figure 8a), if group members do not match at a particular position, various rules may be used to generate a consensus sequence, which is then stored in a consensus BAM file (also known as a read-disintegration BAM file) as a single consensus alignment sequence read for each group of reads (family of reads corresponding to parent fragments). The bases most frequently found within a group may be retained as the consensus (simple majority vote). Alternatively, the consensus may be improved using a quality score (weighted scoring). The resulting consensus sequence can then be processed by any conventional raw read-based variant caller.More generally, as will be apparent to those skilled in the art of NGS bioinformatics workflows, a consensus sequencing approach suitable as an intermediate step of folding aligned reads into a single polynucleotide sequence prior to variant calling 153 can be used in combination with the proposed numerical code tag, as well as the freely available prior art methods outlined in Xu, e.g., the MAGERI bioinformatics workflow (MAGERI: Computational pipeline for molecular-barcoded targeted resequencing), Shugay et al., PLoS Comput. Biol. 2017 May; 13(5)), or various commercially available genome data analysis workflows, e.g., the Illumina Read Collapsing step (https: / / support.illumina.com / help / BaseSpace_App_UMI_Error_Correction_OLH_1000000035906 / Content / Source / Informatics / Apps / Read_Collapsing_appUMI.htm).

[0081] However, the conventional consensus sequencing approach described above has several limitations, which can be overcome by using more advanced genome data analysis workflows based on sophisticated statistical modeling, such as data-driven methods derived from signal processing or machine learning algorithms. Figure 8b) shows stochastic sequencing as another embodiment of the consensus sequencing approach. In stochastic sequencing, instead of generating a consensus BAM file in an intermediate step between alignment and variant calling, the genome data analyzer 150 can directly use raw fragment-tagged alignment files to supply a raw set of aligned reads as input to the statistical variant caller.

[0082] Instead of relying on consensus sequences obtained by heuristic rules (e.g., majority voting), this class of variant callers relies on statistical models that describe how instrument artifacts affect reads belonging to the same or different families (or groups). These statistical models may incorporate knowledge such as:

[0083] In the presence of a mutant DNA molecule, the variant is supported by all reads generated from the two strands of that mutant molecule; • Sequencing errors can occur frequently, but they can occur independently across reads belonging to the same family or reads belonging to different families; PCR errors are infrequent but can affect multiple reads within the same family, and rarely occur on both the positive and negative strands of the same DNA molecule.

[0084] By analyzing the entire set of reads within such a probabilistic framework, it becomes possible, for example, to calculate the posterior probability of the desired mutant allele frequency. This posterior probability can then be used to generate variant calls (e.g., when the probability of the mutant allele frequency is greater than 0 and the probability p is greater than a threshold) and to quantify their confidence levels (i.e., the probability that the signal was generated by actual deformation rather than instrument noise).

[0085] One recently disclosed example of such a statistical variant caller is the freely available, standalone SmCounter2 statistical variant caller, which takes aligned reads as input to calculate variant probabilities according to an error model based on a beta distribution for background error rates and a beta-binomial distribution for the number of non-referenced UMI outliers (smCounter2: an accurate low-frequency variant caller for targeted sequencing data with unique molecular identifiers, Xu et al., Bioinformatics, Vol. 35(8), April 2019). smCounter2 accepts both raw UMI-tagged BAM files and consensus BAM files as input. In the proposed workflow, instead of UMI tags, the UMI-tagged BAM files may also contain numerical code tags of our proposed method, i.e., a pair of numbers corresponding to the measured lengths of variable adapters attached to each end of fragments concatenated according to the proposed wet laboratory method. Similar to SmCounter2, various variant callers from data-driven modeling-based commercial workflows, such as Sophia Genetics Data-Driven Medicine software (Sophia DDM), can also be adapted to individually call variants in each group of aligned reads generated from different DNA fragments, based on the proposed numerical code tagging.

[0086] Exemplary experiment Experiment 1 In the first experiment, the inventors confirmed that all proposed adapters, including variable-length spacer sequences as shown in Figure 5b, for example, could be ligated to DNA fragments to generate a library of DNA-adapter products. As shown by the measurements in Figure 9, when the reaction mixture with the proposed adapters is used during the ligation reaction, all spacer sequence adapters can be ligated to DNA fragments and are represented almost identically in the final DNA library.

[0087] Experiment 2 In the second experiment, we confirmed that the library of DNA-adapter products generated in the first experiment could be sequenced on an NGS platform such as an Illumina NextSeq sequencer and decoded by a genome data analyzer 150 such as the Sophia Genetics Data Driven Medicine (Sophia DDM) bioinformatics platform. Each spacer sequence can be sequenced from the raw FASTQ file by the Sophia Genetics Data Driven Medicine genome data analyzer 150. The resulting reads show a predicted sequence that begins with a truncated spacer subsequence and ends with a constant termination subsequence TS. Figure 10 shows that even with the longest adapters (which are prone to base call errors), the bioinformatics workflow can assign more than 93% of the reads to the predicted spacer sequences. On average, it is possible to properly identify the spacer sequences for approximately 95% of the reads (and thus measure their variable length to form numerical code tags).

[0088] Experiment 3 In the third experiment, the inventors compared a genome analysis bioinformatics workflow (Sophia Genetics DDMv5) that ignored numerical code tagging, i.e., grouping reads obtained for specific genome locations based only on out-of-alignment start and end positions (Figure 11a), with the same genome analysis bioinformatics workflow (Figure 11b) that was further adapted to group reads obtained for specific genome locations based on an NGS data viewer and additional fragment tag information consisting of out-of-alignment start and end positions as well as the numerical code of the proposed method, which consists of measured variable adapter spacer sequence lengths at both ends of the fragment.

[0089] As shown in Figures 11a) and 11b), the NGS data viewer highlights the genomic location 1100 of the heterozygous SNP. In a group of PCR replicas that do not distinguish between origin fragments, theoretically, all reads should either display the SNP (and the downstream variant caller portion 153 measures variant fraction = 1) or not display it (and the downstream variant caller portion 153 measures variant fraction = 0). However, in our actual experiments, as can be seen in Figure 11a), simply grouping the reads by their start and end information does not accurately identify the group of PCR replicas because the actual variant fraction of the SNP is different from 0 or 1. This indicates that these groups contain DNA fragments derived from at least two original DNA fragments. These original DNA fragments differed at the location of the SNP, but they were grouped together because they shared the same start and end locations. In contrast, as can be seen in Figure 11b), by adding the proposed numerical codes as tags, it becomes possible to resolve these collisions by further subdividing and clustering read groups of PCR replicas having the same start and end positions into subsets derived from the same parent fragment according to those numerical codes in the BAM file. In these subgroups, the variant fraction of the SNP can then be measured as either 0 or 1, as expected, by the downstream variant caller section 153, thus demonstrating that the proposed numerical codes, combined with the start and end positions of the reads, enable the distinction of PCR replicas from colliding molecules.

[0090] Experiment 4 Motivation – As will be apparent to those skilled in the art, variant calling in low-variant allele fractions (VAFs) is limited by sequencing errors and library preparation artifacts. Strategies to improve the analytical performance of NGS assays lie in utilizing the information provided by PCR replication for variant calling. Conventional solutions attempt to accurately identify groups of PCR replicas, for example, by mapping their locations to identify the PCR replicas. However, due to the presence of multiple shear points (and therefore mapping locations), this may not be sufficient to distinguish all original DNA molecules. Thus, exogenous molecular barcodes have been introduced to provide additional information for the identification of groups of PCR replicas. However, there is currently no consensus on the best industrial approach for generating such exogenous barcodes, and many conventional solutions require the use of expensive library generation solutions, most of which are designed primarily for use in consensus sequencing workflows and do not benefit from recent advances in stochastic variant calling solutions. In contrast, the proposed variable-length DNA-adapter constructs aim to further improve the sensitivity and specificity of low-frequency variant detection by simultaneously facilitating both the exogenous identification of fragments and their efficient probabilistic genomic analysis. This is demonstrated by dedicated experiments, as detailed herein.

[0091] Nucleosomal DNA from six cell lines was spiked into the nucleosomal DNA of a seventh cell line at different ratios, generating three samples with a series of single nucleotide variants (SNVs) at the following variant allele frequencies: 0.5–4%, 0.25–2%, and 0.1–0.8%.

[0092] Target Library Preparation - Whole-genome libraries were prepared in double strands from 25 ng of each DNA mixture using the SOPHiA GENETICS Library Preparation Kit, with some modifications to the manufacturer's instructions. Briefly, after end repair and A-tailing, the DNA fragments of each sample were attached to a standard unbarcoded adapter or (Figure 5b)(L TS =3, L S =9, generating 10 different DNA adapters ranging in length from 3 to 12 nucleotides. These were ligated to one of a series of variable-length adapters containing variable-length spacer sequences, as shown in ( ). The library was then amplified using indexed Illumina-compatible primers. The whole-genome library was captured using the SOPHiA GENETICS capture protocol and SOPHiA GENETICS catalog panel (installation area: 56Kb), which cover 23 SNVs present in the DNA mixture.

[0093] Data Analysis - Libraries from variable-length adapter construction experiments were first preprocessed. The locations of the constant subsequences at the start points of forward and reverse reads were determined. Then, using the lengths of the adapters present on both sides of each DNA fragment, combinatorial codes were generated to be appended to the read headers before trimming the variable-length adapter sequences. Next, all library reads were aligned to the genome using a BWA-MEM aligner. PCR replica groups were identified using fragment mapping locations and the aforementioned combinatorial codes. Variant calling was performed using either probabilistic sequencing or dual consensus sequencing. For probabilistic sequencing, the posterior probability of PCR replica groups generated from molecules with SNVs was calculated and used to assign a quality score to each identified PCR replica group.

[0094] result Figure 12 shows that the proposed variable-length adapter facilitates the detection of rate variants in artificial nucleosomal DNA. Figure 12a) shows the variant calling results for three samples (25 ng of DNA input) analyzed with replicas containing 23 SNVs in three different VAF ranges (Sample 1: 0.5–4%; Sample 2: 0.25–2%; Sample 3: 0.1–0.8%) when a standard adapter of the prior art is used. Figure 12b) shows the variant calling results obtained when the proposed variable-length adapter is used. Of the 144 SNVs tested in this experiment, only 107 were detected when using the standard adapter. Sensitivity was improved when using the SLA library, with 123 variants being called. Figure 12c) further compares ROC curves showing the performance of variant calling in terms of true positive rate (TPR) to false positive rate (FP) when using probabilistic sequencing (dark gray) or dual consensus sequencing (light gray), respectively, for samples with variants having a VAF in the range of 0.1–0.8% and processed using a conventional standard adapter. Figure 12d) further compares ROC curves showing the performance of variant calling in terms of true positive rate (TPR) to false positive rate (FP) pair when using probabilistic sequencing (dark gray) or dual consensus sequencing (light gray), respectively, for samples with variants having a VAF in the range of 0.1–0.8% and processed using the proposed variable-length adapter.

[0095] Advantages of the proposed method Therefore, the proposed method requires only the ligation of a few predetermined variable-length adapters to generate a library of DNA-adapter products suitable for various downstream NGS workflows, and facilitates NGS bioinformatics identification of variants even with small amounts of DNA input.

[0096] As is evident to those skilled in the field of high-throughput sequencing data processing, in genome analysis workflows, trimming of adapter sequences during read preprocessing must be precise, as over-trimming can lead to loss of sequencing coverage and under-trimming can introduce sequencing artifacts. Prior art variable-length adapters that do not have a constant termination partial sequence signal (TS) cannot identify the boundary between the barcode and the start of the inserted DNA fragment. As a result, they typically require trimming above the full adapter length Lmax, reducing their applicability.

[0097] Furthermore, the synthesis of adapters is costly, and in routine clinical workflows, it is preferable to use the number of barcodes necessary to resolve inconsistencies for specific applications. When using a limited number of barcodes, it is important that they are represented uniformly in the final library; otherwise, the number of effective barcode combinations decreases and may no longer be sufficient. By having a constant termination subsequence (TS) at the end of each barcode, the linking of sequence-specific biases is prevented, thus enabling more uniform barcode usage.

[0098] Furthermore, depending on the actual sequencing technology, for example, with Illumina sequencers, a base imbalance in the first sequencing cycle can degrade sequencing quality. This can be problematic when using a limited number of random barcodes. High sequencing quality can be maintained by using a predetermined set of variable-length spacer sequences that can be designed to have a balanced base composition in each sequencing cycle.

Claims

1. A method for generating a library of DNA adapter products from at least two DNA fragments to facilitate the identification of DNA fragments in an analysis workflow of high-throughput sequencing data after amplification and sequencing, A step of generating a pool of adapters in the form of double-stranded or partially double-stranded polynucleotides, each spacer sequence having spacer sequences at the double-stranded ends, wherein each spacer sequence is linked to a variable spacer subsequence of length L ＴＳ It includes a steady termination subarrangement TS of the length L ＴＳ It consists of three or more nucleotides, and the variable spacer subarrangement has a length of L Ｓ It is cleaved from a common, fixed, predetermined nucleotide sequence (S) having the nucleotides, where the length L S is 5 ≤ ​​L Ｓ The number of nucleotides is ≤20, and the length of each spacer sequence is between LTS and Lmax, where Lmax is the sum of the lengths LTS and LS. A step of producing a first DNA adapter product by ligating first and second adapters from the adapter pool to the respective ends of a first double-stranded DNA fragment in a reaction mixture, wherein the constant termination sub-sequences TS of the first and second adapters are located between the variable spacer sub-sequence of the adapter and the first double-stranded DNA fragment, respectively, and as a result, the first DNA adapter product is formed by ligating the first and second adapter spacer sequences (SS) １ , SS ２ The length of each of the (L) １ , L ２ Characterized by a numerical code formed by ), where L1 and L2 may be the same or different, process, In the same reaction mixture, a step of ligating the third and fourth adapters from the pool of said adapters to each end of a second double-stranded DNA fragment to produce a second DNA adapter product, wherein upon this ligation, the constant termination partial sequences TS of each of said third and fourth adapters are each located between the variable spacer partial sequence of the adapter and said second double-stranded DNA fragment, such that the second DNA adapter product is characterized by a numerical code formed by the respective lengths (L ３ , SS ４ ) of the third and fourth adapter spacer sequences (SS ３ , L ４ ), where L3 and L4 may be the same or different, step, Includes, A method for determining whether the first DNA adapter product and the second DNA adapter product correspond to different fragments, based on the fact that the stationary termination subsequence TS is positioned between the double-stranded DNA fragment and the variable spacer subsequence, and on the characterization of the numerical codes of the first DNA adapter product and the second DNA adapter product.

2. The method according to claim 1, wherein the constant termination subsequence TS exhibits two or more mismatches among the certain predetermined nucleotide sequences (S) compared with the variable spacer subsequence.

3. The method according to claim 1 or 2, wherein the variable spacer subarrangement is obtained by cleaving a certain predetermined nucleotide sequence (S) from its 5' end.

4. The method according to claim 1 or 2, wherein the variable spacer subarrangement is obtained by cleaving a certain predetermined nucleotide sequence (S) from its 3' end.

5. The method according to any one of claims 1 to 4, wherein the constant termination partial sequence TS is a triplet nucleotide ending with a T overhang to facilitate linking to the DNA fragment.

6. The method according to any one of claims 1 to 4, wherein the constant termination subsequence TS is a quadroplet nucleotide ending with a T overhang to facilitate linking to the DNA fragment.

7. A step of amplifying a DNA adapter product to generate PCR replicas suitable for high-throughput sequencing, A step of sequencing the PCR replicas with a high-throughput sequencer to generate raw sequencing reads, The method according to any one of claims 1 to 6, further comprising:

8. Each sequencing lead R ｎ Regarding To generate trimmed sequencing reads, from the start of the read, L ｍａｘ The process of trimming nucleotides, The process of recording trimmed sequencing reads into a pre-processed sequencing read file. The process of aligning trimmed sequencing reads to a reference genome from a pre-processed sequencing read file so that each trimmed read is mapped to its start and end positions. The method according to claim 7, further comprising:

9. Each sequencing lead R ｎ Regarding The first L of the sequencing lead R n ｍａｘ Search for stationary termination subsequences TS in nucleotides, Spacer array SS Ｒｎ Length L ｎ The sequencing lead R ｎ A process of measuring the starting point as a function of the number of nucleotides separated from the ending point of the stationary termination subsequence TS, From the starting point of the lead, at least L ｎ The process of trimming nucleotides to produce trimmed sequencing reads, Measured length L ｎ and the process of recording the trimmed sequencing reads into a pre-processed sequencing read file, The process of aligning trimmed sequencing reads to a reference genome from a pre-processed sequencing read file so that each trimmed read is mapped to its start and end positions. The method according to claim 7, further comprising:

10. Each Sequencing Lead R ｎ Regarding L ｍａｘ The method according to claim 9, wherein a number of nucleotides equal to the number of nucleotides is trimmed from the start of the read.

11. Each Sequencing Lead R ｎ Regarding the spacer array, the number of L corresponding to the measured length is ｎ The method according to claim 9, wherein the nucleotides are trimmed from the start of the read.

12. Sequencing generates paired-end reads, which are aligned to the same start and end positions relative to the reference genome sequence read direction. Paired-end reads with the same numerical code pair for the measured spacer sequence length (L1, L2) are grouped as sequencing reads arising from the same original double-stranded DNA fragment. If the numerical code pair for the measured spacer sequence length (L1, L2) is an F1R2 oriented paired-end read, then {L ｎ（ｆｏｒｗａｒｄ） , L ｍ（ｒｅｖｅｒｓｅ） } In the case of paired-end leads having F2R1 orientation {L ｎ（ｒｅｖｅｒｓｅ） , L ｍ（ｆｏｒｗａｒｄ） The method according to any one of claims 9, 10, or 11, further comprising the step of further subdividing those paired-end reads into two subgroups according to their strand origin, which of the two strands of the original double-stranded DNA fragment they originate from, given by}.

13. The method according to claim 12, further comprising the steps of folding each read group that shares the same start position, end position and numeric code into a consensus array of their parent fragments, and identifying the variant of this parent fragment into the folded consensus array using a variant calling method.

14. The method according to claim 12, further comprising identifying the probability of a variant for their parent fragment using a statistical variant calling method for each read group that shares the same start position, end position and numerical code.

15. A multiplex high-throughput sequencing genomic analysis method for identifying genomic variants in at least two patient samples from a pool of samples, comprising generating a library of DNA adapter products using the method described in claim 1, wherein the library of DNA adapter products differs between samples.

16. The method according to claim 15, wherein the library of DNA adapter products differs between samples by a certain termination subsequence TS within each sample.

17. The method according to claim 15 or 16, wherein the library of DNA adapter products differs between samples by a certain predetermined nucleotide sequence (S) in each sample that is cleaved to obtain a variable spacer subsequence.

Images