Systems and methods for determining nucleic acid sequences based on mutant sequence reads

Abstract

Described herein are methods and systems for determining the sequence of a nucleic acid template by removing mutations present in mutant sequence reads. In some embodiments, the methods and systems include the steps of comparing unmutated sequence reads to mutated sequence reads of a nucleic acid template, and determining the most probable sequence of the nucleic acid template and a score of accuracy per base associated with the probability that a base matches the nucleic acid template, thereby determining the sequence of the nucleic acid template.

Description

[0001] Incorporation by reference into any priority application

[0002] Any and all applications claiming foreign or domestic priority identified in the Application Data Sheet filed with the present application are hereby incorporated by reference pursuant to 37 CFR 1.57.

[0003] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 489,550, filed on March 10, 2023, which is hereby incorporated by reference in its entirety. Background Art Technical Field

[0004] The present disclosure relates generally to the field of nucleic acid sequencing. More specifically, the present disclosure relates to the field of determining the sequence of a nucleic acid template by removing mutations present in mutant sequence reads.

[0005] describe

[0006] Conventional nucleotide sequence analysis techniques, known as "error correction" or "polishing," are used to correct errors that are inadvertently introduced into the sequencing process due to noisy or fallible measurements of the primary sequence signal. However, such techniques have been criticized as being ineffective, particularly with repetitive regions. In some cases, such methods are employed to generate an "average sequence" of copies of a repetitive region, rather than an accurate representation of any particular copy of a nucleic acid template containing the repetitive region. Summary of the invention

[0007] The methods disclosed herein each have several aspects, no one of which is solely responsible for its desired properties. Without limiting the scope of the claims, some salient features will now be briefly discussed. Many other embodiments are also contemplated, including embodiments with fewer, additional and / or different components, steps, features, purposes, benefits and advantages. The various components, aspects and steps may also be arranged and sequenced in different ways. After considering this discussion, particularly after reading the section entitled "Specific Embodiments," it will be understood how the features of the systems and methods disclosed herein provide advantages over other known systems and methods.

[0008] Described herein is a computer-implemented method for determining the sequence of a nucleic acid template by removing mutations present in a mutant sequence read. In some embodiments, the method includes comparing an unmutated sequence read of the nucleic acid template with a mutant sequence read, wherein the mutant sequence read comprises one or more mutations, and determining the most likely sequence of the nucleic acid template, and for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template; thereby determining the sequence of the nucleic acid template.

[0009] In some embodiments, the mutant sequence reads have a mutant nucleic acid template. In some embodiments, one or more mutations include mutations introduced during sample preparation. In some embodiments, one or more mutations include mutations randomly introduced during sample preparation by mutagenesis. In some embodiments, one or more mutations include mutations that were unintentionally introduced. In some embodiments, the nucleic acid template is longer than the mutant sequence read and the unmutated sequence read.

[0010] In some embodiments, determining the most likely sequence of the nucleic acid template comprises determining the most likely base for each position associated with the alignment of the unmutated sequence reads with the mutant sequence reads. In some embodiments, determining the most likely sequence of the nucleic acid template comprises determining the most likely sequence for a subset of positions associated with the alignment of the unmutated sequence reads with the mutant sequence reads.

[0011] In some embodiments, determining the most probable sequence of the nucleic acid template and, for each base of the most probable sequence, a per-base accuracy score comprises summing, on a site-by-site basis, the probability of each base at each site being associated with a non-mutated sequence read.

[0012] In some embodiments, determining the most probable sequence of the nucleic acid template and, for each base of the most probable sequence, the per-base accuracy score includes incorporating information related to the probability of a mutation at each site of a mutated sequence read.

[0013] In some embodiments, determining the most probable sequence of the nucleic acid template and, for each base of the most probable sequence, a per-base accuracy score includes incorporating information describing the probability of errors associated with the sample preparation or sequencing process.

[0014] In some embodiments, determining the most probable sequence for the nucleic acid template and, for each base of the most probable sequence, the per-base accuracy score comprises determining the most probable base for the position based on a statistically inferred probability of the position for a mutant sequence read not aligning to a non-mutated sequence read.

[0015] In some embodiments, determining the most probable sequence for the nucleic acid template and, for each base of the most probable sequence, a per-base accuracy score includes incorporating information describing variations in mutation rates between mutant sequence reads or unequal mutation rates for adenine, cytosine, guanosine, and thymine.

[0016] In some embodiments, the nucleic acid template is derived from a nucleic acid sample that comprises two or more copies of a repetitive region of the nucleic acid template.

[0017] In some embodiments, determining the most probable sequence of the nucleic acid template and, for each base of the most probable sequence, the per-base accuracy score comprises estimating the probability that the non-mutated sequence read is derived from a copy of the same repeat region as the mutated sequence read to which the non-mutated sequence read is aligned.

[0018] In some embodiments, determining the most likely sequence of the nucleic acid template and, for each base of the most likely sequence, a per-base accuracy score comprises applying an optimization process, wherein the optimization process iteratively updates an estimate of the probability that the non-mutated sequence read is derived from the same copy of the mutant sequence read aligned with the non-mutated sequence read, while updating the inference of the most likely base. In some embodiments, the optimization process comprises applying an expectation maximization method. In some embodiments, the optimization process incorporates information describing variations in read length.

[0019] In some embodiments, the contribution of each non-mutated sequence read to the probability calculation of the most likely base is weighted by a weighting scheme. In some embodiments, the weighting scheme is related to the sequence identity of the non-mutated sequence read and the current iteration estimate of the sequence of the nucleic acid template.

[0020] In some embodiments, determining the most likely sequence of the nucleic acid template and, for each base of the most likely sequence, a per-base accuracy score comprises selecting a subset of non-mutated sequence reads among non-mutated sequence reads derived from different copies of the repeat region. In some embodiments, selecting a subset of non-mutated sequence reads comprises scoring alignments of non-mutated sequence reads to mutated sequence reads and selecting the alignment with the highest score.

[0021] In some embodiments, the method includes determining the sequence of a copy of the repeat region of the nucleic acid template. In some embodiments, the two or more copies of the repeat region include two or more haplotypes, and the method includes determining the sequence of the haplotype of the nucleic acid sample.

[0022] In some embodiments, the per-base accuracy score comprises a per-base quality value (QV).In some embodiments, the method further comprises determining an overall accuracy score associated with the most likely sequence.

[0023] In some embodiments, the method further comprises, before aligning the non-mutated sequence reads of the nucleic acid template with the mutated sequence reads, assembling the non-mutated sequence reads. In some embodiments, aligning the non-mutated sequence reads of the nucleic acid template with the mutated sequence reads comprises aligning the mutated sequence reads with components of the non-mutated sequence reads.

[0024] In some embodiments, the method further comprises, before aligning the non-mutated sequence reads of the nucleic acid template with the mutated sequence reads, assembling the mutated sequence reads. In some embodiments, aligning the non-mutated sequence reads of the nucleic acid template with the mutated sequence reads comprises aligning the non-mutated sequence reads with components of the mutated sequence reads.

[0025] In some embodiments, the method further comprises, prior to aligning the non-mutated sequence reads to the mutated sequence reads of the nucleic acid template, aligning the non-mutated sequence reads or the mutated sequence reads to a reference sequence.

[0026] In some embodiments, the method further comprises, before comparing the unmutated sequence reads of the nucleic acid template with the mutated sequence reads, assembling the mutated sequence reads, generating a consensus sequence of the assembled mutated sequence reads, and comparing the unmutated sequence reads with the consensus sequence. In some embodiments, comparing the unmutated sequence reads of the nucleic acid template with the mutated sequence reads comprises replacing the consensus sequence with the mutated sequence reads in the comparison of the unmutated sequence reads with the consensus sequence. In some embodiments, assembling the mutated sequence reads comprises comparing the mutated sequence reads with a reference sequence.

[0027] In some embodiments, the method further comprises sequencing the non-mutated nucleic acid template molecules and the mutated nucleic acid template molecules to obtain non-mutated sequence reads and mutated sequence reads.

[0028] In some embodiments, the method includes aligning unmutated sequence reads and mutated sequence reads to a reference sequence; determining an initial rendering sequence by determining a most likely base for each position associated with the alignment of the unmutated sequence reads to the mutated sequence reads, in combination with information describing variations in mutation rates between the mutated sequence reads or unequal mutation rates for adenine, cytosine, guanosine, and thymine; and determining an additional rendering sequence by comparing k-mers from the initial rendering sequence to k-mers from the unmutated sequence reads.

[0029] In some embodiments, the method includes aligning unmutated sequence reads and mutated sequence reads to a reference sequence; realigning unmutated sequence reads having mapping positions that directly overlap with mutated sequence reads with the mutated sequence reads; scoring the alignments of the unmutated sequence reads with the mutated sequence reads and selecting the highest scoring alignment; and determining a rendered sequence by determining the most likely base for each position associated with the alignment of the unmutated sequence reads with the mutated sequence reads.

[0030] In some embodiments, the method includes determining a rendered sequence by determining a most probable base for each site associated with an alignment of a non-mutated sequence read and a mutant sequence read, and applying an expectation-maximization (EM) optimization process, wherein the EM optimization process iteratively updates an estimate of the probability that the non-mutated sequence read is derived from the same copy of the mutant sequence read to which the non-mutated sequence read is aligned, while updating an inference of the most probable base.

[0031] In some embodiments, the method includes generating a consensus sequence of assembled mutant sequence reads; aligning the non-mutated sequence reads with the consensus sequence; replacing the consensus sequence with the mutant sequence reads in the alignment of the non-mutated sequence reads with the consensus sequence; and determining the rendered sequence by: determining the most likely base for each site associated with the alignment of the non-mutated sequence reads and the mutant sequence reads, and applying an expectation-maximization (EM) optimization process, wherein the EM optimization process iteratively updates an estimate of the probability that the non-mutated sequence read is derived from the same copy of the mutant sequence read to which the non-mutated sequence read is aligned, while updating the inference of the most likely base.

[0032] In another aspect, disclosed herein is a system comprising a processor configured to perform a method. In some embodiments, the method comprises comparing a non-mutated sequence read of a nucleic acid template with a mutated sequence read of a mutated nucleic acid template, wherein the mutated sequence read comprises one or more mutations, and determining a most likely sequence of the nucleic acid template, and for each base of the most likely sequence, a per-base accuracy score associated with a probability that the base matches the nucleic acid template; thereby determining a sequence of the nucleic acid template.

[0033] In another aspect, computer readable media are disclosed herein. In some embodiments, the computer readable medium comprises instructions that, when executed by a processor, perform a method comprising the following steps: comparing a non-mutated sequence read of a nucleic acid template with a mutated sequence read of a mutated nucleic acid template, wherein the mutated sequence read comprises one or more mutations, and determining the most likely sequence of the nucleic acid template, and for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template; thereby determining the sequence of the nucleic acid template. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Embodiments of methods for determining the sequence of a nucleic acid template by removing mutations present in mutant sequence reads are shown.

[0035] Figure 2 An embodiment of a system comprising a processor configured to perform a method for determining a sequence of a nucleic acid template by removing mutations present in mutant sequence reads is shown.

[0036] Figure 3 An example of using mutated sequence reads and non-mutated sequence reads to determine the sequence of a nucleic acid template is shown.

[0037] Figure 4 An example of determining the sequence at a site containing a heterozygous transition SNP is shown.

[0038] Figure 5 An example of determining the sequence of a real human genome dataset is shown. DETAILED DESCRIPTION

[0039] All patents, applications, published applications, and other publications mentioned herein are incorporated herein by reference in their entirety. If a term or phrase is used herein in a manner that is contrary to or inconsistent with the definition set forth in the patents, applications, published applications, and other publications incorporated herein by reference, the use herein prevails over the definition incorporated herein by reference.

[0040] Some embodiments relate to systems and methods for determining the nucleotide sequence of a nucleic acid template. In some cases, sequencing a nucleic acid template may involve introducing random mutations into the template so that the template can be fragmented into smaller sequences that are more easily passed through the next generation sequencing process. The combination of random mutations can help to rebuild the sequence of a mutant nucleic acid template by mutual comparison of mutant fragments. However, in some cases, the sequence of an initial unmutated template molecule is desired. Embodiments relate to methods for determining the unmutated sequence of an initial nucleic acid template.

[0041] An embodiment of determining the unmutated sequence of the initial nucleic acid template can be related to fragmenting the unmutated nucleic acid template and determining the sequence of the unmutated fragment. The system and method can then read the unmutated sequence from the nucleic acid template and compare it with the mutated sequence of the nucleic acid template to determine the most likely sequence of the unmutated nucleic acid template. The process can include, for each base of the most likely unmutated nucleic acid template, determining the per base accuracy score related to the probability of the base matching the nucleic acid template. By using this per base accuracy score, the most likely sequence of the unmutated nucleic acid template can be determined. This will be described in more detail in the following sections.

[0042] definition

[0043] As used herein, "alignment" refers to the arrangement of two or more nucleic acid sequences to identify regions of similarity.

[0044] A "nucleic acid template" can be any nucleic acid molecule that a user wants to sequence. A "nucleic acid template" can be single-stranded, or can be part of a double-stranded complex. If the nucleic acid template is composed of deoxyribonucleotides, it can form part of a double-stranded DNA complex. In this case, one chain (e.g., a coding chain) can be considered to be a nucleic acid template, and the other chain is a nucleic acid molecule complementary to the nucleic acid template. A nucleic acid template can be a DNA molecule corresponding to a gene, can contain introns, can be an intergenic region, can be an intragenic region, can be a genomic region spanning multiple genes, or can actually be the entire genome of an organism. A nucleic acid template can include repetitive sequences, for example, variable number tandem repeats (VNTR) or short tandem repeats (STR). A nucleic acid template can be derived from a nucleic acid sample.

[0045] "Nucleic acid sample" includes any sample of nucleic acid. A nucleic acid sample can be a sample of nucleic acid derived from a human, such as a sample extracted from a skin swab of a human patient. A nucleic acid sample can be a sample of nucleic acid from a non-human. A nucleic acid sample can be derived from a sample such as from a water source. Such a sample may contain hundreds, thousands, millions or even billions of nucleic acid template molecules. A nucleic acid sample can include a single genome, or a nucleic acid sample can include multiple non-identical genomes (such as from different organisms or different species or strains). A nucleic acid sample can contain multiple haplotypes from a single organism.

[0046] As used herein, "repeat region" refers to a region within a nucleic acid template, wherein the nucleotide sequence of the region is replicated and present in at least one other instance in a nucleic acid sample. In other words, a nucleic acid sample may have two or more copies of a repeat region of a nucleic acid template. In some embodiments, the two copies of the repeat region may be different, for example, containing a single base difference. In some embodiments, the repeat region contains a repetitive sequence (e.g., VNTR or STR), such as where the repetitive sequences are adjacent to each other or in series. For example, a nucleic acid sample may include a nucleic acid template having a repeat region including a repetitive sequence, and the nucleic acid sample may include another copy of the repeat region, which carries a repetitive sequence elsewhere (e.g., on another chromosome, or at a different position on the same chromosome). As another example, in some embodiments, the repeat region may include the full length of a chromosome. In this embodiment, a "copy" of the repeat region will be another copy of a chromosome present in the nucleic acid sample, such as a second haplotype.

[0047] "Mutated nucleic acid template" refers to a nucleic acid template that contains one or more mutations, such as a mutant base, compared to the original sequence of the nucleic acid template. In other words, a mutant nucleic acid template refers to a mutant copy of a nucleic acid template.

[0048] As used herein, "mutation" refers to a change in a nucleic acid sequence. For example, a mutation includes the displacement of one nucleotide base pair by another. The mutation can be a transition mutation (from A to G, G to A, C to T, or T to C). In another embodiment, the mutation is a transversion mutation. In some embodiments, the mutation is intentionally introduced into the nucleic acid via a sample processing step that randomly introduces the mutation into the nucleic acid molecule. In some embodiments, the mutation is unintentionally introduced, such as due to an error or unexpected consequence during sample preparation or library preparation. In some embodiments, the mutation is caused by an unintentional error during sequencing of the nucleic acid template, such as a base call error.

[0049] As used herein, "mutation sequence reads", also referred to herein as "dream" or "dream sequence", refer to sequence reads containing mutations compared to a nucleic acid template. In some embodiments, a mutation sequence read is a sequence read of a mutant nucleic acid template. In some embodiments, a mutation sequence read is obtained by sequencing a region of a mutant nucleic acid template. In some embodiments, a mutation sequence read is obtained by sequencing a region of a nucleic acid template and contains mutations due to sequencing errors. The mutation sequence read may have a length that is smaller than the length of the nucleic acid template. In some embodiments, the mutation sequence read has a length of about 100nt to 600nt, such as about 150nt or about 300nt. In some embodiments, the mutation sequence reads can be assembled by aligning the mutation sequence reads. Any method known in the art can be used for such an alignment, such as the method described in US20210174905A1.

[0050] As used herein, "unmutated sequence read" refers to a sequence read of a nucleic acid template. In some embodiments, the unmutated sequence read is obtained by sequencing a region of the nucleic acid template. The unmutated sequence read may have a length that is less than the length of the nucleic acid template, such as a length of about 100 nt to 600 nt, such as about 150 nt or about 300 nt.

[0051] "Rendering" refers to the process of determining the most likely sequence of a nucleic acid template using an alignment of unmutated sequence reads and mutated sequence reads. In some embodiments, rendering includes determining the most likely sequence of a nucleic acid template by replacing mutations in a mutated sequence read with the most likely base.

[0052] "Most probable base" refers to the nucleic acid base calculated to have the highest probability of matching with the nucleic acid template (in some cases, a specific copy of the nucleic acid template) from which the mutant sequence read is derived. For example, the most probable base can be a maximum a posteriori base as described herein.

[0053] "Most likely sequence" refers to a nucleic acid sequence calculated to have the highest probability of matching a nucleic acid template (in some cases, a specific copy of a nucleic acid template) from which the mutant sequence reads are ultimately derived (such as an unmutated nucleic acid template that existed before the introduction of the mutation). Thus, in some embodiments, the most likely sequence does not include (is statistically predicted to not include) the mutation. Thus, in some embodiments, the most likely sequence can be generated by statistically removing the mutation from the mutant sequence reads to estimate the sequence of the (unmutated) nucleic acid template.

[0054] Method for determining the sequence of a nucleic acid template using mutated sequence reads and non-mutated sequence reads

[0055] Embodiments relate to methods for determining the sequence of a nucleic acid template molecule using a sample processing step that intentionally introduces random mutations into a copy of the sample nucleic acid. For example, methods for introducing random mutations into a target nucleic acid molecule are described in U.S. Patent Application Publication No. US20210010008A1. In some embodiments, the mutations are Complete Long Read Sequencing Sample Preparation Chemical Introduction. The introduction of such random mutations can be used for sequence assembly, for example, as described in U.S. Patent Application Publication No. US20210174905A1. For example, when the nucleic acid template includes a repetitive region, such as when the nucleic acid sample includes two or more non-identical copies of the repetitive region (e.g., haplotypes), the generation of mutant copies of the nucleic acid template may be particularly useful for performing nucleic acid sequencing.

[0056] Therefore, some embodiments of the systems and methods disclosed herein include the step of obtaining one or more nucleic acid samples (including obtaining a nucleic acid sample separated into two nucleic acid samples, and including two nucleic acid samples from the same subject). Some embodiments include randomly introducing mutations into nucleic acid molecules in one or more nucleic acid samples to produce mutant nucleic acid samples. Some embodiments include keeping the nucleic acid sample unmutated by not randomly introducing mutations into nucleic acid molecules in the nucleic acid sample. Some embodiments herein also include sequencing nucleic acid molecules (such as nucleic acid templates) of (unmutated) nucleic acid samples to produce unmutated sequence reads. Some embodiments disclosed herein also include sequencing nucleic acid molecules (such as mutant nucleic acid templates) of mutant nucleic acid samples to produce mutant sequence reads. The sequencing of mutant and unmutated nucleic acid samples can be performed simultaneously or sequentially. Sequencing can be achieved by any method known to those skilled in the art, including next generation sequencing (NGS).

[0057] Although mutant sequence reads can be used to determine the sequence of the nucleic acid template, it may also be desirable to determine the sequence of the initial nucleic acid template molecule as it existed before the random mutations were introduced. To achieve this, the mutant sequence reads can be aligned with the non-mutated sequence reads to determine the sequence of the nucleic acid template in a process referred to herein as "rendering".

[0058] However, in some cases, determining the alignment of a mutant sequence read to a non-mutated sequence read may be ambiguous. For example, when the nucleic acid template has multiple copies of a repetitive region, such as a repetitive region containing a repetitive sequence, and / or when the nucleic acid sample contains two or more haplotypes of the nucleic acid template, the alignment of the non-mutated sequence read to the mutant sequence read may be ambiguous because the mutant sequence read may be aligned at many different positions on the nucleic acid template.

[0059] The present disclosure relates to systems and methods for accurately predicting the sequence of a nucleic acid template given one or more mutant sequence reads and one or more non-mutated sequence reads from the same nucleic acid sample. In some embodiments, the method involves aligning the mutant sequence reads and the non-mutated sequence reads, and statistically determining the most likely sequence of the nucleic acid template and a per-base accuracy score corresponding to each base of the most likely sequence.

[0060] Various computer-implemented methods for determining the correct sequence of a nucleic acid template by removing mutations present in a mutant sequence read are described herein. In some embodiments, a mutant sequence read comprises one or more nucleic acid mutations. In some embodiments, a mutant sequence read has a mutant nucleic acid template. In some embodiments, one or more mutations comprise mutations introduced during sample preparation. In some embodiments, one or more mutations are randomly introduced during sample preparation by mutagenesis. In some embodiments, mutations comprise intentionally introduced nucleic acid mutations. The method for introducing such mutations may be any method known in the art. In some embodiments, mutations comprise substitution mutations. In some embodiments, mutations comprise conversion mutations.

[0061] In some embodiments, the nucleic acid template is longer than the mutant sequence read and / or the unmutated sequence read. For example, in some embodiments, the unmutated sequence read and the mutant sequence read are greater than 10bp, greater than 100bp, less than 200bp, less than 300bp, less than 400bp, less than 500bp, less than 600bp, less than 700bp, less than 1,000bp, less than 3,000bp, or within the range constructed by any of the aforementioned values. In some embodiments, the length of the unmutated sequence read and the mutant sequence read is about 300bp. In some embodiments, the length of the unmutated sequence read and the mutant sequence read is about 150bp.

[0062] In some embodiments, the methods described herein are used to remove mutations that have been inadvertently introduced into the sequencing process, such as by noisy or error-prone measurements of the primary sequence signal. In such embodiments, the methods described herein can be considered an improved method of error correction or polishing.

[0063] Alignment with reference sequences and / or consensus sequences

[0064] Some embodiments include utilizing Figure 1 The method shown removes mutations present in mutant sequence reads to determine a system or method for the sequence of a nucleic acid template. As shown, the method begins at step S100, in which sequence reads, such as unmutated sequence reads and / or mutant sequence reads, are compared with a reference sequence.

[0065] Can be compared by methods known to those skilled in the art.The reference sequence can be derived from a nucleic acid sample different from the nucleic acid template, or can be from the same nucleic acid sample.In some embodiments, the reference sequence is from an organism species identical to the sample nucleic acid template.In some embodiments, the reference sequence is a publicly available reference sequence.

[0066] In some embodiments, the system and method further comprises step S110 of assembling the mutant sequence reads, generating a consensus sequence of the assembled mutant sequence reads, and aligning the non-mutated sequence reads with the consensus sequence.

[0067] The work of sequence analysis by mutagenesis previously developed the technology of accurately calling consensus sequence from mutant sequence collection, such as disclosed in Keith JM, Adams P, Bryant D, Cochran DAE, Lala GH, Mitchelson KR. Algorithms for sequence analysis via mutagenesis. Bioinformatics. 2004 05; 20 (15): 2401-2410. doi: 10.1093 / bioinformatics / bth258. In order to determine the most probable sequence of nucleic acid template, the previous method will build the multiple comparison of mutant sequence reading, or via reference sequence mapping, or via de novo multiple comparison method, called consensus sequence, and the consensus sequence is copied to each site in the corresponding site read by the mutant sequence of comparison, so as to determine the most probable sequence of nucleic acid template. When using reference sequence, it can be independently constructed, or read and constructed by unmutated sequence, or read and constructed by mutant sequence.

[0068] In some embodiments, instead of directly using the consensus sequence to determine the most likely base or most likely sequence, the consensus sequence of the mutant sequence read is used to improve the accuracy of the comparison of the unmutated sequence read with the mutant sequence read. In short, in step S110, for example, the consensus sequence of the mutant sequence read is called using the method disclosed by Keith et al. (2004) to generate the consensus sequence of the mutant sequence read. Further in step S110, the consensus sequence is then used as the mapping target of the unmutated sequence read. Further in step S110, the comparison on the consensus sequence is then translated back to the original (non-shared) mutant sequence read, and the comparison of the unmutated sequence read with the original mutant sequence read is used in the method for determining the sequence described herein.

[0069] One advantage of this approach over directly aligning the unmutated sequence reads with the initial mutated sequence reads is that the consensus sequence may have a much higher sequence similarity with the unmutated sequence reads derived from the same haplotype. As further described herein, the nucleic acid sample from which the nucleic acid template is derived may include multiple copies of a repeat region, such as multiple haplotypes, which may make the alignment of the unmutated sequence reads with the mutated sequence reads ambiguous or incorrect. Higher sequence identity means that mapping and alignment are more accurate, calculations are faster, and may contain fewer non-mutated sequence reads that are aligned with the mutated sequence reads from another copy of the repeat region (such as another haplotype). One advantage of this approach over using a consensus sequence method to determine the most likely sequence is that using unmutated sequence reads can provide much higher confidence and accuracy in the resulting base calls.

[0070] Alignment of mutated and non-mutated sequence reads

[0071] In some embodiments, the methods and systems described herein further include a step S120 of aligning the non-mutated sequence reads of the nucleic acid template with the mutated sequence reads. The mechanism for generating such an alignment can be any method known to those skilled in the art.

[0072] In some embodiments, after the step of generating a consensus sequence S110, a step S120 of aligning the non-mutated sequence reads with the mutated sequence reads is performed by replacing the consensus sequence with the mutated sequence reads in the alignment of the non-mutated sequence reads with the consensus sequence. In some embodiments, a consensus sequence is not generated, and the non-mutated sequence reads and the mutated sequence reads are aligned by mapping each other to each other by methods known to those skilled in the art.

[0073] In some embodiments, the alignment of the mutant sequence reads to the unmutated sequence reads can include about 30 times coverage of the unmutated sequence reads. In some embodiments, the alignment includes about 10 times coverage, about 15 times coverage, about 20 times coverage, about 30 times coverage, about 40 times coverage, about 50 times coverage, about 60 times coverage or higher of the unmutated sequence reads, or a range constructed by any of the foregoing values. In some embodiments, the alignment includes coverage as described above, for example, 15 times coverage of a single haplotype of the nucleic acid template sample, with the unmutated sequence reads.

[0074] In some embodiments, the method further comprises, prior to the alignment in step S120, assembling the non-mutated sequence reads to generate components of the non-mutated sequence reads. In some embodiments, the non-mutated sequence reads are assembled using a reference sequence. In some embodiments, the step S120 of aligning the non-mutated sequence reads with the mutated sequence reads may comprise aligning the mutated sequence reads with the components of the non-mutated sequence reads.

[0075] In some embodiments, prior to step S120 of aligning the non-mutated sequence reads with the mutated sequence reads, the method includes assembling the mutated sequence reads to generate components of the mutated sequence reads. In some embodiments, the mutated sequence reads are assembled by using a reference sequence. In some embodiments, step S120 of aligning the non-mutated sequence reads with the mutated sequence reads may include aligning the non-mutated sequence reads with components of the mutated sequence reads.

[0076] Methods for making multiple copies of a repeated region

[0077] In some embodiments, the nucleic acid template is derived from a nucleic acid sample of two or more copies of a repeat region comprising the nucleic acid template. For example, the nucleic acid sample may include two or more haplotypes. For example, the nucleic acid sample may include a repeat region, which is included in another position on the nucleic acid molecule in the nucleic acid sample (such as, on another chromosome or in a different region of the same chromosome, in other words, in a different position in the genome present in the nucleic acid sample) with a copy of the sequence (such as a repetitive sequence). In some embodiments, each copy of the repeat region has a slight difference from each other, so that the nucleic acid sample includes two or more different copies of the repeat region.

[0078] A challenge faced by methods for determining a target nucleic acid sequence using mutant sequence reads and non-mutated sequence reads is the uncertainty of whether any given non-mutated sequence read covering a repetitive region has been derived from the "same copy" of a mutant sequence read covering the repetitive region. "Same copy" can be conceptualized as meaning that an error-free non-mutated sequence read will be 100% identical to the homologous portion of the (unknown) error-free mutant sequence read (minus the mutation) to which it has been aligned. This challenge can be particularly difficult when the sequence reads cover long repetitive sequence regions of a nucleic acid template.

[0079] A further problem arising from mutagenesis (and error-prone sequencing) is that a particular change in a repeat region can alter the sequence of one repeat region to make it more similar to another copy of the repeat region that may be present elsewhere in the nucleic acid sample (such as on another chromosome or in a different region of the same chromosome, or on another haplotype.) This will also lead to ambiguity in the alignment of non-mutated sequence reads with mutated sequence reads.

[0080] When the above method is applied in its simplest form, a non-mutated sequence read covering one copy of the repeat region may be aligned to a mutated sequence read covering another copy of the repeat region, thereby introducing errors. Aligning a non-mutated sequence read to an incorrect copy of the repeat region in a mutated sequence read may prevent accurate identification of mutation sites within the mutated sequence read and may result in a large number of low-confidence (low QV) base calls (e.g., bases with low per-base accuracy scores) in the most likely sequence.

[0081] When the nucleic acid sample from which the nucleic acid template is derived includes two or more copies of a repeat region, various exemplary methods for determining the most likely sequence of the nucleic acid template are described in further detail below. For example, the method for solving this problem may include step S130 of applying an optimization process as further described herein, or a step of selecting a subset of non-mutated sequence reads as further described herein.

[0082] In some embodiments, the method includes determining the sequence of a copy of a repetitive region of a nucleic acid template. As a non-limiting example, the systems and methods described herein can be used to determine the sequence of a copy of a repetitive region in a nucleic acid template, wherein the repetitive region (such as a repetitive region containing a repetitive sequence) has another non-identical copy at another location in the nucleic acid sample (such as elsewhere in the genome). In some embodiments, two or more copies of the repetitive region include two or more haplotypes, and the method includes determining the sequence of the haplotype of the nucleic acid sample. As a non-limiting example, a nucleic acid sample from a human being can include two haplotypes. Embodiments of the systems and methods disclosed herein can advantageously determine the sequence of one or more haplotypes.

[0083] Application of optimization process

[0084] After the process has aligned the non-mutated sequence reads and the mutant sequence reads at step S120, the system and method can move to step S130 of applying the optimization process (which includes steps S140, S150, and S160, which can be repeated). In some embodiments, in step S130, the optimization process iteratively updates the estimate of the probability that the non-mutated sequence read is derived from the same copy of the mutant sequence read aligned to the non-mutated sequence read, while updating the inference of the most likely base.

[0085] As previously described, a challenge faced by rendering and related polishing or error correction techniques is the handling of repetitive regions, where similar but not identical copies of repetitive regions are present in different abundances in a nucleic acid sample. For this reason, it may be advantageous to determine that a non-mutated sequence read has been derived from a copy of the same repetitive region as the mutant sequence read to which it has been aligned (or to estimate its probability), and to incorporate this estimated information into the rendering process so that sequence reads that may have been derived from another copy of the repetitive region contribute less to the final result. Various exemplary methods are described below, such as step S130 of applying an optimization model. In some embodiments, a combination of any of these methods is used to estimate whether a non-mutated sequence read has been derived from a copy of the same repetitive region as a mutant sequence read.

[0086] This can be described by reference to the following non-limiting examples. In one example, there are two similar but not identical repeat regions, r1 and r2, wherein the r2 sequence appears in 10 positions in the genome in the nucleic acid sample, while the r1 sequence only appears in a single position. Alternatively, in another example, in a metagenomic environment, r2 appears in a species or strain that is 10 times more abundant than the species or strain containing r1 in the nucleic acid sample. In these two examples, the comparison of the unmutated sequence reads and the mutant sequence reads will generally produce the comparison of the unmutated sequence reads from both r1 and r2 and the mutant sequence reads from r1 and r2. Problems arise when determining the sequence of the nucleic acid template containing r1, because compared with the unmutated sequence reads from r1, the unmutated sequence reads from r2 more than 10 times are compared with the mutant sequence reads of r1 (incorrectly). In the absence of a technique to select the correct non-mutated sequence reads (those derived from the same copy of the repeat region r1), when r1 is rendered, the non-mutated sequence reads from r2 will dominate the signal and cause the predicted most likely sequence of r1 to bear the r2 sequence as a byproduct of rendering. This phenomenon has been reported in the academic community when using standard error correction or assembly polishing techniques, resulting in poor accuracy for repeat regions. However, techniques that can weight the contribution of non-mutated sequence reads to rendering (such as those described herein) have the ability to select the correct reads for rendering in these cases.

[0087] As a non-limiting example, the method may include step S140, estimating the probability of each unmutated sequence read being compared with one or more mutant sequence reads, that is, the unmutated sequence read is generated from a copy (such as the same haplotype) of the same repeat region as the one or more mutant sequence reads compared with the unmutated sequence reads. In some embodiments, step S140 includes estimating the probability of generating an unmutated sequence read for comparison with one or more mutant sequence reads at a given site by sequencing a nucleic acid molecule, the nucleic acid molecule containing a copy of the same repeat region as the one or more mutant sequence reads compared with the unmutated sequence reads. In some embodiments, the unmutated sequence reads cover the repeat region. For example, the probability that the unmutated sequence reads are derived from the same haplotype as the mutant sequence reads compared with the unmutated sequence reads can be estimated in step S140. Various statistical models can be used in step S140 to estimate the probability that the unmutated sequence reads are derived from the same copy of the repeat region as the mutant sequence reads compared with the unmutated sequence reads. Therefore, the systems and methods described herein may include step S140, estimating the probability that the unmutated sequence reads are derived from the same copy of the repeat region as the mutant sequence reads compared with the unmutated sequence reads.

[0088] In some embodiments, the method further comprises step S150, wherein the contribution of each non-mutated sequence read to the probability calculation of the most probable base is weighted by a weighting scheme. For example, in step S150, when statistically inferring the most probable base, non-mutated sequence reads that are aligned with mutant sequence reads may be given unequal weights. For example, if for a particular site in the alignment of non-mutated sequence reads with mutant sequence reads, some non-mutated sequence reads have base "G" and some have base "A", then when statistically inferring the most probable base for that site, non-mutated sequence reads with base "G" may be given more weight.

[0089] In some embodiments, in step S150, the weighting scheme is related to the current iteration estimate of the sequence identity of the unmutated sequence read and the sequence of the nucleic acid template. For example, in step S150, the weighting scheme can be related to the estimated probability that the unmutated sequence reads the copy of the same repeat region (such as the same haplotype and / or position in the genome) that is derived from the mutant sequence read that is compared with the unmutated sequence read. For example, in some embodiments, each pairwise comparison of the unmutated sequence read and the mutant sequence read has a weight, and the weight is based on the estimated probability that the unmutated sequence reads and the mutant sequence reads the same copy from the repeat region. This estimated probability can be updated iteratively as described above. In step S150, the weighting scheme can also be related to the current iteration estimate of the most likely sequence of the nucleic acid template, wherein the most likely sequence is iteratively updated as described herein. Therefore, in some embodiments, in step S150, the weighting scheme can be based on the current estimated probability of the sequence identity of the unmutated sequence read, and based on the most likely sequence currently predicted by the nucleic acid template, iteratively update the weighted contribution of each unmutated sequence read compared with one or more mutant sequence reads at a given site.

[0090] In step S150, the contribution of the unmutated sequence reading can be weighted based on the current estimated probability, such as the estimated probability based on the current iteration of the optimization process. Therefore, when determining the most likely sequence, the estimated probability of the unmutated sequence reading can be factored in. In step S160, the process proceeds to update the inference of the most likely base to the site. In some embodiments, the inference of the most likely base includes the calculation of the probability distribution of the base (such as A, C, G, T) at the site. In some embodiments, the most likely base is the base with the highest statistical probability of matching with the nucleic acid template at the site of the base. In some embodiments, in step S160, the method can introduce the weighted contribution of the unmutated sequence reading in updating the inference of the most likely base to the site. In some embodiments, in step S160, the inference of the most likely base is updated for each site in the subset of the site.

[0091] In some embodiments, in step S140, the updated probability can be re-estimated using the updated inference of the most likely base of step S160. In step S150, the contribution of the non-mutated sequence read can be weighted using the updated estimated probability, and in step S160, the inference of the most likely base, such as for the second site or a subset of sites, is updated. In some embodiments, the process is repeated until the change in the resulting most likely sequence or probability distribution over the base is below a convergence threshold.

[0092] The advantage of using such an optimization process as described in step S130 is that the optimization process is iteratively performed toward a specific copy of the repetitive region of the nucleic acid template in the nucleic acid template sample, rather than determining an average sequence across multiple copies. By proceeding toward a specific copy of the repetitive region, the method allows the most likely sequence of a specific copy of the repetitive region in the nucleic acid template to be determined, rather than generating an "average" of the repetitive region across different copies of the repetitive region in the nucleic acid sample. Therefore, in some embodiments, the methods and systems described herein include determining the sequence of a copy of the repetitive region (such as one of two or more non-identical copies). In some embodiments, when two or more copies of the repetitive region in the nucleic acid sample include two or more haplotypes, the method includes determining the sequence of a haplotype of the nucleic acid sample (such as one of two or more non-identical haplotypes).

[0093] As an illustrative example, in some embodiments, the optimization process of step S130 includes applying the expectation maximization method.Expectation maximization (EM) is a statistical inference method, which is usually used when there is a correlation between the parameters in the probability model, and this correlation prevents the direct use of analytical techniques to calculate expectations, mean values, variances, etc. As a non-limiting example, according to the EM iterative method, in step S130, the comparison of the mutant sequence reading and the unmutated sequence reading is used as input. The site can be evaluated on the basis of each site to find a site with high confidence for the most likely base of the site. In step S160, the inference of the most likely base can be updated by calculating the probability distribution on the base of the site (such as on A, C, G, T). Then, the probability distribution on the base (or the set of such distribution in all comparison sites) can be used to re-estimate the probability of the unmutated sequence reading comparison in step S140. For the region of high confidence, the updated comparison can be re-evaluated. Now the site with sufficiently high confidence can be used to determine the most likely base of the site. This process can be repeated until a convergence threshold is reached.

[0094] In some embodiments, in step S130, the optimization process incorporates information describing variations in read lengths. This may be problematic in real (non-simulated) data sets where sequence reads may be of unequal lengths due to the nature of sequencing chemistry and preliminary data analysis steps (such as read trimming), and information about read lengths may typically be factored into the optimization process. Therefore, information about such variations in sequence read lengths may be incorporated into the optimization process. For example, in some embodiments, the EM method uses a pseudo-likelihood defined on a fixed data dimension (read length).

[0095] Summarize unmutated sequence reads and combine information

[0096] In some embodiments, the process of calculating the inference of the most likely base includes a step of summarizing the probability of each base at the site associated with the unmutated sequence read. For example, in some embodiments, the step S160 of updating the inference of the most likely base includes a step S161, summarizing the probability of each base at each site associated with the unmutated sequence read on a site-by-site basis. For example, in some embodiments, for each site associated with the unmutated sequence read, the probability of a nucleotide A, C, G, or T at a specific position in the unmutated sequence read is summarized. In some embodiments, in Example 1, the summary can be expressed as formula (3).

[0097] In some embodiments, the method of calculating the inference of the most likely base includes a step of combining information related to the probability of mutation at the site, the probability of error (such as unintentionally introduced mutation or sequencing error) and / or the mutation rate and local sequence composition changes. For example, after the site-by-site analysis of step S161, the method may include step S162, combining information related to the probability of mutation at each site of the mutant sequence read. For example, in the case of intentionally introduced mutations, the probability of mutation at each site in the mutant sequence read can be statistically inferred based on knowledge of the mutation rate of the mutagenesis process used in the sample preparation recipe.

[0098] In some embodiments, step S162 includes incorporating information describing the probability of errors associated with the sample preparation or sequencing process. For example, in some embodiments, in step S162, the method includes considering changes other than intentionally introduced mutations (such as mutations that are unintentionally introduced). In some embodiments, in step S162, the method takes into account changes in the sequence, such as those introduced by using PCR during sample preparation and sequencing, or other sources of sequencing errors. In some embodiments, in step S162, such other changes can be considered in conjunction with the probability of intentionally introduced mutations at the site.

[0099] Depending on the mutagenesis process, sample preparation, library preparation, and sequencing methods used, the probability of intentionally introduced mutations may be different from the probability of unintentional changes in the sequence in the mutant sequence reads. For example, in step S162, the method can consider at least four different possible scenarios occurring at each site in the mutant sequence reads: 1) intentional mutations (such as transition mutations) and no useless mutations (such as polymerase errors); 2) unintentional mutations and no intentional mutations; 3) both intentional mutations and unintentional mutations; 4) neither intentional mutations nor unintentional mutations.

[0100] In some embodiments, the method includes the probability of statistical inference of the mutation at the site for the site that is not compared with the unmutated sequence read to determine the most likely base. For example, in some embodiments, some sites in the mutant sequence read may not be compared with any unmutated sequence read (may not be mapped). In some embodiments, in the absence of information about the site read from the unmutated sequence, it is possible to read only based on the mutant sequence covering the site, combined with the probability of statistical inference of mutation at the site, determine the most likely base for the site. In some embodiments, the accuracy score per base of the site will also be based on the same process.

[0101] In some embodiments, in step S162, the process includes incorporating information describing changes in mutation rates between mutant sequence reads or unequal mutation rates of adenine, cytosine, guanosine, and thymine. In some embodiments, in step S162, the method may include introducing information describing changes in mutation rates between mutant sequence reads when statistically predicting the most likely sequence. For example, although the mutagenesis process may have an average mutation rate, such as 5%, some mutant sequence reads may randomly have more or less mutations than the average, such as 3% or 8%. Similarly, in step S162, when statistically predicting the most likely sequence, the process may include incorporating information describing unequal mutation rates of adenine, cytosine, guanosine, and thymine. For example, the mutagenesis process may produce different mutation rates between A, C, G, and T nucleotide sites, and this information may be incorporated into the methods described herein.

[0102] Determine the most likely sequence and per-base accuracy score

[0103] In some embodiments, after the optimization process of step S130, the method further comprises step S170 of determining: i) the most likely sequence of the nucleic acid template, and ii) for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template.

[0104] In some embodiments, the per-base accuracy score includes a per-base quality value (QV), also referred to as a "PHRED" score. In some embodiments, the per-base accuracy score estimates the statistical confidence in the prediction of each base. In some embodiments, the method also includes determining an overall accuracy score associated with the most probable sequence. In some embodiments, the overall accuracy score is determined based on the per-base accuracy score of each base in the bases in the most probable sequence. For example, the overall accuracy score can be an average or median of the per-base accuracy scores, or another combination of per-base accuracy scores.

[0105] In some embodiments, the step S170 of determining the most probable sequence of the nucleic acid template includes determining the most probable base for each site associated with the comparison of the unmutated sequence reading and the mutant sequence reading, as described herein. In some embodiments, the method includes determining the most probable sequence on a site-by-site basis. For example, the method may include statistically inferring the most probable base of a site (position in the nucleic acid sequence) by analyzing the comparison of the mutant sequence reading and the unmutated sequence reading at the site, and then proceeding to another site, etc. For example, in some embodiments, the method includes determining the maximum a posteriori base for each site, and the most probable sequence includes one or more maximum a posteriori bases. In some embodiments, determining the most probable base for each site includes calculating the inference of the most probable base to the site, such as by calculating the probability distribution on the base (such as A, C, G, T) at the site. In some embodiments, the most probable sequence includes one or more most probable bases determined by, for example, applying an optimization process in step S130.

[0106] In some embodiments, the step S170 of determining the most probable sequence of the nucleic acid template includes determining the most probable sequence for a subset of sites associated with the comparison of the unmutated sequence reads with the mutated sequence reads. Therefore, in some embodiments, the method includes determining the most probable sequence on a subset-by-subset basis. As used herein, a subset of sites refers to a group of sites contained within the most probable sequence. In some embodiments, the subset of sites is less than the total number of sites within the most probable sequence.

[0107] In some embodiments, the most likely sequence and each base accuracy score are stored in a file on a computer storage medium (such as a computer hard drive, e.g., a spinning disk drive or a solid state drive). In some embodiments, the file is stored in the format of a BAM, SAM, CRAM, or VCF file.

[0108] Select a subset of unmutated sequence reads

[0109] In some embodiments, the system and method of the present invention include a method for selecting a subset of unmutated sequence reads in unmutated sequence reads of different copies derived from a repeat region. In some embodiments, the subset of unmutated sequence reads selected includes a subset of unmutated sequence reads selected, and the unmutated sequence reads are compared with one or more mutant sequence reads at the site to contribute to the statistical inference of the most likely base of the nucleic acid template at the site. In some embodiments, other unselected unmutated sequence reads compared with the mutant sequence reads at the site are excluded from the statistical inference of the most likely base of the nucleic acid template at the site or contribute less to the statistical inference. In some embodiments, after the most likely base of the site is statistically inferred as described herein, the process of selecting a subset of unmutated sequencing reads is repeated at a second site. In some embodiments, a subset of unmutated sequences is selected in unmutated sequence reads of different copies of a repeat region derived from a nucleic acid sample.

[0110] In some embodiments, selecting a subset of unmutated sequence reads includes scoring the alignment of unmutated sequence reads with mutated sequence reads, and selecting the alignment with the highest score. In some embodiments, the scoring scheme is specific to the alignment of unmutated sequence reads with mutated sequence reads at a specific site (such as a site predicting the most likely base). For example, in some embodiments, the alignment of unmutated sequence reads with mutated sequence reads is scored using an alignment scoring scheme, and the best "n" alignments are selected according to the alignment scoring scheme.

[0111] In embodiments of the present invention, various scoring schemes can be used to score the comparison. As a non-limiting example, in some embodiments, a standard scoring scheme is used to score the comparison. In some embodiments, the scoring scheme can be a 4x4 permutation matrix and an affine gap penalty score. In some embodiments, the scoring scheme is the probability generated by using a method such as a pair-HMM to integrate the comparison. In some embodiments, another method known in the art is used to score the comparison of the unmutated sequence read and the mutant sequence read.

[0112] As a non-limiting example, after the comparison is scored, an iterative process is applied to select the highest score for each site. For example, for a specific site, the highest score of "n" comparisons of the specific site is selected, where "n" refers to the determined number of comparisons. For example, "n" can be a number such as 20, so that the first 20 comparisons are selected. "n" can be 2, 5, 10, 15, 20, 25, 30, 40, 50, or a range constructed by any of the above values. In some embodiments, "n" is related to the coverage depth of the unmutated sequence reads used in the comparison. For example, in some embodiments, "n" corresponds to 50% of the coverage depth, where if the coverage is 40x, "n" is 20.

[0113] In some embodiments, after the comparison that selects the highest score for this site, determine the most likely base (such as the maximum a posteriori base) as described herein. This most likely base can be incorporated into the most likely sequence. In some embodiments, use the current estimate of the most likely sequence, repeat the process to extra sites. For example, the process can be repeated until, as a result of the repetition, less than the specified score of the base change. Can determine every base accuracy score for each base, and further describe above. In some embodiments, determine the most likely sequence and every base accuracy score thus.

[0114] system

[0115] On the other hand, systems are disclosed herein. For example, a system may include one or more processors configured to perform any method described herein that can be implemented by a computer. In some embodiments, a system including a processor is disclosed herein that is configured to perform a method including the following items: a) comparing an unmutated sequence read of a nucleic acid template with a mutant sequence read, wherein the mutant sequence read contains one or more mutations, and b) determining: i) the most likely sequence of the nucleic acid template, and ii) for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template; thereby determining the sequence of the nucleic acid template.

[0116] Figure 2 is a block diagram of an exemplary computing device 200 that may be used in conjunction with an exemplary sequencing system. Computing device 200 may be configured to determine a sequence of a nucleic acid template by removing mutations present in mutant sequence reads. Figure 2 The general architecture of the depicted computing device 200 includes an arrangement of computer hardware and software components. The computing device 200 may include Figure 2 Those more (or fewer) elements are shown. However, in order to provide an enabling disclosure, it is not necessary to show all of these generally conventional elements.

[0117] As shown, computing device 200 includes processing unit 210, network interface 220, computer readable medium drive 230, input / output device interface 240, display 250 and input device 260, all of which can communicate with each other through a communication bus. Network interface 220 can provide connection with one or more networks or computing systems. Therefore, processing unit 210 can receive information and instructions from other computing systems or services via the network. Processing unit 210 can also communicate with memory 270 via input / output device interface 240, and also provide output information for optional display 250. Input / output device interface 240 can also accept input from optional input device 260 (such as keyboard, mouse, digital pen, microphone, touch screen, gesture recognition system, voice recognition system, game pad, accelerometer, gyroscope or other input device).

[0118] Memory 270 may contain computer program instructions (grouped into modules or components in some embodiments) that processing unit 210 executes to implement one or more embodiments. Memory 270 typically includes RAM, ROM, and / or other persistent, secondary, or non-transitory computer-readable media. Memory 270 may store an operating system 272 that provides computer program instructions for use by processing unit 210 in general management and operation of computing device 200. Memory 270 may also include computer program instructions and other information for implementing aspects of the present disclosure.

[0119] For example, in one embodiment, the memory 270 includes a rendering processing module 274 for determining the sequence of the nucleic acid template by removing mutations present in the mutant sequence read. The rendering processing module 274 can perform the method disclosed herein, including referring to Figure 1 The process described. In addition, the memory 270 may include a data memory 290 and / or one or more other data memories, or communicate with the data memory and / or one or more other data memories, and the data memory and / or one or more other data memories store one or more inputs, one or more outputs and / or one or more results (including intermediate results) of determining the sequence of the nucleic acid template by removing the mutation present in the mutant sequence reading according to the present disclosure. Such data may include, for example, unmutated sequence readings, mutant sequence readings, mutation probability information, error probability information, variable mutation rate information, and most likely sequence and accuracy scores per base.

[0120] In some embodiments, the systems and methods disclosed herein may relate to methods for transferring or distributing certain sequence data analysis features and sequence data storage information to a cloud computing environment or a cloud-based network. User interactions with sequencing data, genomic data, or other types of biological data may be mediated via a central hub that stores and controls access to various interactions with the data. In some embodiments, a cloud computing environment may also provide sharing of medical protocols, analytical methods, libraries, sequence data, and distributed processing for sequencing, analysis, and reporting. In some embodiments, a cloud computing environment facilitates user modification or annotation of sequence data. In some embodiments, the systems and methods may be implemented in a computer browser, on demand, or online.

[0121] In some embodiments, software written for performing the methods as described herein is stored in some form of computer-readable media, such as a memory, CD-ROM, DVD-ROM, memory stick, flash drive, hard drive, SSD hard drive, server, mainframe storage system, etc.

[0122] In some embodiments, the method can be written in any programming language in various suitable programming languages, such as compiled languages ​​such as C, C#, C++, Fortran and Java. Other programming languages ​​can be scripting languages, such as Perl, MatLab, SAS, SPSS, Python, Ruby, Pascal, Delphi, R and PHP. In some embodiments, the method is written in C, C#, C++, Fortran, Java, Perl, R, Java or Python. In some embodiments, the method can be a stand-alone application with data input and data display modules. Alternatively, the method can be a computer software product and can include such a class, wherein the distributed object includes an application containing a computing method as described herein.

[0123] In some embodiments, the method can be incorporated into existing data analysis software (such as data analysis software found on a sequencing instrument). Software including computer-implemented methods as described herein is installed directly onto a computer system, or is maintained indirectly on a computer-readable medium and loaded onto a computer system as needed. Further, the method can be located on a computer remote from the location where the data is generated, such as software found on a server or the like maintained in another location relative to the location where the data is generated (such as provided by a third-party service provider).

[0124] The measuring instrument, desktop computer, laptop computer or server may contain a processor operatively communicating with an accessible memory, and the accessible memory includes instructions for realizing the system and method. In some embodiments, a desktop computer or laptop computer operatively communicates with one or more computer-readable storage media or devices and / or output devices. The measuring instrument, desktop computer and laptop computer can operate under many different computer-based operating languages, such as the operating languages ​​used by the computer system based on Apple or the computer system based on PC. The measuring instrument, desktop computer and / or laptop computer and / or server system can also provide a computer interface for creating or modifying experimental definitions and / or conditions, viewing data results and monitoring experimental processes. In some embodiments, the output device can be a graphical user interface, such as a computer monitor or computer screen, a printer, a handheld device, such as a personal digital assistant (i.e., PDA, blackberry, iPhone), a tablet computer (e.g., iPAD), a hard disk drive, a server, a memory stick, a flash drive, etc.

[0125] Computer readable storage device or medium can be any device such as server, mainframe, supercomputer, tape system, etc. In some embodiments, storage device can be located at the site of the position of measuring instrument, such as adjacent to or next to measuring instrument. For example, relative to measuring instrument, storage device can be located in the same room, in the same building, in adjacent buildings, on the same floor in a building, on different floors in a building, etc. In some embodiments, storage device can be located outside the measuring instrument site or away from the measuring instrument. For example, relative to measuring instrument, storage device can be located in different parts of a city, different cities, different states, different countries, etc. In the embodiment where storage device is located away from the measuring instrument, communication between one or more of measuring instrument and desktop computer, laptop computer or server is usually connected by the Internet (wirelessly or by access point using network cable). In some embodiments, storage device can be maintained and managed by individuals or entities directly associated with measuring instrument, while in other embodiments, storage device can be maintained and managed by third parties, usually at the location of individuals or entities associated with measuring instrument at the far end. In the embodiment as described herein, output device can be any device for visualizing data.

[0126] The measuring instrument, desktop computer, laptop computer and / or server system itself can be used to store and / or retrieve a computer-implemented software program including a computer code for performing and implementing the calculation method as described herein, data used in implementing the calculation method, etc. One or more of the measuring instrument, desktop computer, laptop computer and / or server may include one or more computer-readable storage media, which are used to store and / or retrieve a computer-implemented software program including a computer code for performing and implementing the calculation method as described herein, data used in implementing the calculation method, etc. Computer-readable storage media may include, but are not limited to, one or more of a hard drive, an SSD hard drive, a CD-ROM drive, a DVD-ROM drive, a floppy disk, a tape, a flash memory stick or a card, etc. Further, a network including the Internet can be a computer-readable storage medium. In some embodiments, a computer-readable storage medium refers to a computing resource storage device that can be accessed by a computer network through a corporate network provided by the Internet or a service provider, rather than, for example, a local desktop computer or laptop computer at a location away from the measuring instrument.

[0127] In some embodiments, computer-readable storage media for storing and / or retrieving computer-implemented software programs containing computer code for performing and implementing computational methods as described herein, data for use in implementing computational methods, etc., are operated and maintained by a service provider that operatively communicates with the assay instrument, desktop computer, laptop computer, and / or server system via an Internet connection or a network connection.

[0128] In some embodiments, the hardware platform for providing a computing environment includes a processor (i.e., a CPU), wherein processor time and memory layout such as a random access memory (i.e., RAM) are system considerations. For example, a smaller computer system provides a cheap, fast processor and a large memory and storage capacity. In some embodiments, a graphics processing unit (GPU) can be used. In some embodiments, the hardware platform for performing a computing method as described herein includes one or more computer systems with one or more processors. In some embodiments, smaller computers are clustered together to create a supercomputer network.

[0129] In some embodiments, the computing methods as described herein are performed on a collection of interconnected or internally connected computer systems (i.e., grid technology) that can run various operating systems in a coordinated manner. For example, the CONDOR framework (University of Wisconsin-Madison) and the systems available through United Devices are examples of coordinating multiple independent computer systems for the purpose of processing large amounts of data. These systems can provide a Perl interface to submit, monitor and manage large sequence analysis jobs on a cluster through serial or parallel configurations.

[0130] Example

[0131] Some aspects of the embodiments discussed above are further disclosed in detail in the following examples, which are not intended to limit the scope of the present disclosure in any way. It will be appreciated by those skilled in the art that many other embodiments also fall within the scope of the present disclosure, as described above and in the claims. For example, the symbols introduced in some of the examples below assume that the unmutated sequence reads and the mutated sequence reads are compared with the reference genome assembly. However, the methods described below can also be used to directly compare the unmutated sequence reads with the mutated sequence reads without using a reference sequence as an intermediate.

[0132] In the examples below, the following notation applies: "Dream" or "dream sequence" refers to a mutant sequence read that contains mutations that can be removed by rendering.

[0133] Denote the set of mapped “reference reads” (non-mutation sequence reads, such as standard non-mutation shotgun reads from the reference sequence) covering site i in the reference sequence as R i , let r∈R i is the read from the set of reads that aligns to site i in the reference sequence. Denote the nucleotide vector in read r that aligns to site i in the reference sequence as r i . r i is a vector of length 4 indicating one of the four nucleotides (e.g., when A is [1; 0; 0; 0], C is [0; 1; 0; 0]). The probability of calling the base at position i in read r as correct, e.g., the quality score is denoted as q(r i ).

[0134] Assume d i represents a vector of length 4, which describes the probability of A, C, G and T at the position in dream aligned with reference site i, with an indicator vector a of length 4 i, where 1 is in the position corresponding to the A, C, G, T of the dream base. As mentioned above, let q(a i ) represents the probability that the base call at position i of the dream sequence matches the nucleic acid template. i Write as follows:

[0135] d i =a i q(a i )+(1-a i )(1-q(a i ) / 3) (1)

[0136] Example 1 :

[0137] Statistical methods are used to determine the nucleic acid sequence by removing mutations introduced into the mutant sequence reads. The details of the statistical methods are described below.

[0138] Maximum a posteriori base calling and Bayes factor quality scores

[0139] set up:

[0140]

[0141] Vector x i is a non-normalized distribution over the reference read bases at site i and is a regularizing prior / pseudo-count used to enforce equal probability for all bases in the absence of reliable data from mapped reference reads. When normalized, this forms a probability distribution over A, C, G, T at site i:

[0142]

[0143] The prior probability of a mutation at any site is expressed as P(u), where u refers to the mutation event. In typical applications of complete long read (ICLR) sequencing, P(u) will take values ​​in the range of 0.05 to 0.06.

[0144] Assume z i A vector of length 4 representing the unnormalized posterior probabilities of A, C, G, and T at site i. i The value of is defined as follows:

[0145]

[0146] in represents the Hadamard product (element-wise product) of vectors / matrices, where y iis a normalized vector of length 4 describing the probability of A, C, G, and T at that position in the reference sequence, which is calculated as described above from the reference read. M is a 4x4 rotation matrix that rotates the elements of the length 4 vector to apply the transformation mutation:

[0147]

[0148] To render the bases, from z i Select the maximum a posteriori value from:

[0149] d′ i = argmax(z i ) (6)

[0150] where d′ i Represents the dream base (most likely base) rendered at alignment position i.

[0151] For site i, set the quality score, expressed as QV(z i ),equal:

[0152]

[0153] The quality score is a Bayes factor that considers bases that are not argmax(z i The assumption that the ) (molecule) is argmax(z i ) assumptions for comparison.

[0154] Example 2: Extending MAP base calling to polymerase error correction

[0155] The above base calling method is extended to determine the sequence when changes other than transition mutations are taken into account, such as those introduced during sample preparation and sequencing by using PCR, or other sources of sequencing errors. The details of this method are described below.

[0156] To account for non-conversion polymerase errors, assume

[0157]

[0158] and let PI be the prior probability of polymerase error, which is expected to be approximately 5x10 per PCR cycle in a typical ICLR library (32 cycles in total). -5 or 1.6x10 -3 The posterior probability over the base is redefined as:

[0159]

[0160] Intuitively, the above formula integrates the probabilities over four scenarios, reflected in the four terms of the formula. The first term corresponds to a mutation event and no polymerase error, the second term corresponds to a polymerase error and no mutation, the third term corresponds to both a mutation and a polymerase error, and the fourth term corresponds to neither a mutation nor an error occurring. Base calling and QV estimation are performed in the same manner as in formulas (6) and (7) above, but using z i That is, the output base is set to the value corresponding to argmax(z i ) at position i of the nucleotide, and the mass fraction QV(z i ) is calculated as the Bayes factor comparing the MAP base to the other three candidate bases.

[0161] Example 3: Handling of unmapped dream locations

[0162] When a site in a mutated sequence read does not map to a non-mutated sequence read, a statistical method is used to determine the sequence. The details of the statistical method are described below.

[0163] As mentioned above, let z i A vector of length 4 representing the posterior probability of the base at site i. Define φ to be a vector of length four representing the probability that any A, C, G, T base in any dream has been mutated. φ extends the concept of P(u) to represent unequal probabilities of mutations at sites containing A, C, G, T. This number can be reliably calculated by counting only the number of transition differences at all A, C, G, T sites in the alignment of all dream sets to the reference reads, and normalizing by the total number of A, C, G, T sites. Formally, φ is defined as:

[0164]

[0165] where a i and r i is the indicator vector for dream and read as defined above, and A, C, G, T are shorthand for the indicator vector corresponding to each base. Thus, φ is defined as a vector of length 4, where one entry for each of A, C, G, T gives the alignment position a in dream i Compare with the reference reads in the alignment set r i The observed difference in the rate of transition between φ is estimated using a set of one or more aligned dreams, and the resulting estimate of φ is then applied when rendering the aligned dreams. That is, the value of φ is expected to be similar between aligned and unaligned dreams in a particular sample / dataset, because they are all subject to the same mutation process during sample preparation.

[0166] If the dream base is not aligned, then set

[0167]

[0168] where i is reinterpreted as an index into the dream sequence rather than an index into the alignment. The first term incorporates the probability that no mutation has occurred, while the second term handles the case where a mutation has occurred. As described above, the output base at position i is set to argmax(z i ), and QV is set using equation (7). In this case, since there is no information from the reference reads, the maximum a posteriori probability (MAP) estimates of the bases and associated quality scores are driven entirely by data on the overall mutation rate.

[0169] Example 4: Adjustment of local sequence composition and unequal mutation rates

[0170] Statistical methods are used to determine the sequence when taking into account local sequence composition and unequal mutation rates. The details of the statistical methods are described below.

[0171] The variation in local sequence composition and the unequal mutation rates at the four bases A, C, G, T are accounted for by using φ, as defined above in the section (see equation (10)). To this end, z i The definition is further extended as follows:

[0172]

[0173] Example 5: Expectation Maximization

[0174] A technique similar to EM is applied to infer the rendering sequence. The details of this technique are described below.

[0175] The probability of reading r is defined as follows This read exists in r In the comparison:

[0176]

[0177] where Θ takes the usual meaning of the symbol as the aggregate of all model parameters;

[0178] where r i is a vector of length four probabilities of A, C, G, T in read r at site i in the alignment of the read to dream, and ι = {0, 1, 2} denotes the values ​​produced by an iteration of EM. Z(·) is simply the previously introduced formula (12) rewritten as a function with a parameter that can vary across iterations of EM:

[0179]

[0180] The remaining symbols are as defined previously. The value of φ0 is initialized as described above in equation (10). d i,0 The value of is as above for d in formula (1) i Initialized.

[0181] The parameter update for each iteration of EM is then defined as follows. First, x and y are redefined to use the likelihood of each read and indexed by the EM iteration:

[0182]

[0183] The c vector sums the likelihood of the read being used at each site and the y i,ι It is used in the definition of to normalize these values ​​to sum to 1. As above, α is the regularization constant.

[0184] The previously defined Z(·) is then used to identify the most likely base at each site, which is expressed as:

[0185] d′ i,ι = argmax Z(e,φ ι ,d i,ι ,y i,ι ) (17)

[0186] And suppose:

[0187]

[0188] As above, in formula (6), d' is used to represent the bases in the fully resolved nucleotide sequence. d is maintained as the probability distribution over the four possible nucleotides at each site, updated at each iteration of EM. All that remains is to define the value of φ at each iteration, which can be calculated based on d' i,ι To do:

[0189]

[0190] That is, φ ι is defined as a vector of length 4, where d′ i,ι One term for each of A, C, G, T in gives the value of i,ι Alignment positions and alignment sets in The rate at which transition differences are observed between the reference reads in .

[0191] Example 6: Practical considerations for real datasets

[0192] A statistical method was used to adjust the expectation maximization method described above, which assumes equal length reads, so that the likelihood is calculated on data of equal dimensions to account for reads of unequal lengths. Details of the statistical method are described below.

[0193] Denote L as the maximum read length that will be processed. In an exemplary application, L = 150. Alternatively, if paired-end 150nt (150 nucleotides) reads are processed as a single entity, then L = 300. For a particular read, is the number of sites in read r that are aligned with dream, and defines:

[0194]

[0195] That is, q r is the average per-position probability of the aligned sites in the read. r To define the pseudo-likelihood of reading a fixed dimension of r, denoted as

[0196]

[0197] In other words, the average per-site likelihood from the aligned sites is taken as an estimate of the per-site likelihood that would be at any missing site to pad the data dimension in the likelihood function to L sites.

[0198] Finally, the function Δ(·) is defined to calculate the number of nucleotide differences between two sequences as follows:

[0199]

[0200] An exemplary method can be represented using the defined symbols and functions as:

[0201]

[0202] Example 7: Simple simulation of mutations in a dataset

[0203] Rendering mutations in simulated data sets using the statistical methods described in this paper, without sequencing errors.

[0204] For example, Figure 3 An example of rendering a site 340 containing a mutation is shown. In the alignment 310, a mutant sequence read 320 is compared to several aligned non-mutated sequence reads 330, and mismatches are highlighted at site 340, where the mutant sequence read has a "T" at site 340 and the non-mutated sequence read has a "C" at site 340. Figure 3The simplest version of a mathematical model of the mutation and sequencing process is used in the example of . In this example, the unmutated sequence reads 330 are consistent in their support for alternative nucleotides compared to the mutated sequence reads 320, and this difference is explained by the transition mutation. The maximum a posteriori base generated by the model matches the base present in the unmutated sequence read 330, so the rendering process corrects the mutation. Figure 3 In the example of , the simple model outlined in equations (6) and (7) is used. An alignment of the non-mutated sequence read 330 with the mutated sequence read 320 is shown in alignment 310. i ,q(a i ), r i ,q(r i ), α i The values ​​of and P(u) are specified as constants in Table 360. Using the constant values ​​in the formula to calculate d i 、x i ,y i 、z i and QV(z i ) are shown in Table 370. i The MAP base selected is C (corresponding to z i 0.0506), and the resulting PHRED scaled QV score is 17.2, corresponding to an estimated probability of about 0.98 that the rendered base call is correct. In a typical dataset with higher coverage of unmutated data (e.g., a larger number of aligned unmutated sequence reads 330), the QV score will be higher.

[0205] Example 8: SNPs in a simple simulation dataset

[0206] Mutations were rendered in simulated data sets using the statistical methods described herein without sequencing errors.The data set of this example contained at least two haplotypes that were heterozygous for a single nucleotide polymorphism (SNP).

[0207] Figure 4 An example of a rendered site is shown where the unmutated data shows a heterozygous SNP where the two bases involved represent a transition difference. In alignment 410, the mutated sequence read 420 matches one of the two bases present at site 440 in the unmutated sequence read 430, leaving a high degree of uncertainty as to whether site 440 is mutated. In table 470, the MAP estimate produced by the model leaves the base unchanged and assigns it a low QV, corresponding to a probability of the base being correctly called of approximately 0.92. The low confidence in the base call reflects the prior knowledge that approximately 5% of the sites in the sequence are mutated due to the nature of the sequencing chemistry. Figure 4In the example of , the simple model outlined in equations (6) and (7) is used. The alignment of the unmutated sequence reads to the mutated sequence reads is shown in alignment 410. a i ,q(a i ), r i ,q(r i ), α i and P(u) are specified as constants in Table 460. Using the constants in the formula to calculate d i 、x i ,y i 、z i and QV(z i ) are shown in Table 470. i The MAP base selected in is G (corresponding to z i 0.3779), and the resulting PHRED-scaled QV score is 10.9, corresponding to an estimated probability of about 0.92 that the rendered base call is correct.

[0208] Example 9: Accuracy evaluation in simulated datasets

[0209] The accuracy of the rendering method was evaluated on a simulated data set that included a mutant sequence read template, a "reference genome" that simulated the mutant template, and an unmutated "reference read" (unmutated sequence read) also simulated from the reference genome. The advantage of evaluating on a simulated data set is that the correct rendered sequence (the sequence corresponding to the nucleic acid template) is known, and therefore the accuracy of the rendering is explicitly evaluated. The details of this experiment are described below.

[0210] Simulation data generation

[0211] Inputs to the simulation process include the reference genome assembly, target coverage of reference reads, target coverage of long reads, and short read coverage (redundancy) for each long read template.

[0212] The simulation process produces several outputs, including a true set of long template sequences (representing nucleic acid template sequences), and their positions in the reference genome assembly, a set of long template sequences with simulated mutations (representing mutant nucleic acid templates), the positions of the mutations introduced into the long templates, a set of standard short reads (representing non-mutated sequence reads), a set of short reads simulated from mutant long template sequences (representing mutant sequence reads), and information about which long template each mutant short read originated from.

[0213] The simulation system was used to generate three datasets, called "small", "medium" and "large", whose characteristics are listed in Table 1 below. The human simulation used the diploid assembly of HG002, which is not available at NCBI. Small contigs were removed from the assembly before data simulation. During analysis, the human dataset used real reference read sequences instead of simulated reads.

[0214] Table 1

[0215] Dataset name Reference genome Accession Number Genome size Long reads Reference Reading Small Staphylococcus aureus NC_007795.1 2.82Mbp 56.4Mbp 84.6Mbp middle People T2T chr22 CP068256.2 51.3Mbp 153Mbp 1.54Gbp big People HG002 - 3.1 Gbp 89.4Gbp 105Gbp

[0216] All datasets were simulated with long reads drawn from a normal distribution with an average of 6 kbp and a mutation rate of 6%. Mutations were located at positions drawn uniformly at random within each long sequence and consisted strictly of transition mutations.

[0217] Rendering implementation

[0218] Four alternative implementations of the rendering methods were evaluated: initial, minihyb, EM, and EM w / cons. These four methods represent simplifications of the operations of various combinations of the methods presented above. A description of each method follows.

[0219] initial

[0220] In the "initial" method, non-mutated sequence reads and mutated sequence reads are mapped to a reference sequence. The initial stage of rendering is then performed using the formula defined in Example 4: Adjustment for local sequence composition and unequal mutation rates.

[0221] In the subsequent rendering stage, k-mer comparisons are used to further remove any residual errors in the data. Briefly, the k-mer present in each rendered read ("rendered k-mer") is compared to the set of k-mers present in the unmutated sequence read ("unmutated k-mer"). When there is no rendered read k-mer in the set of unmutated k-mers, and the k-mer is adjacent to a rendered k-mer that may be present in the unmutated k-mer, an attempt is made to change the non-existent k-mer so that it may be present in the unmutated k-mer set. The nucleotide in the first k-mer that is not covered by the second k-mer is modified by applying a transition mutation, and if the resulting k-mer can then be present in the set of unmutated k-mers, the change is maintained, otherwise it is restored. This process is repeated on all adjacent k-mers in each rendered read until changes that increase the number of rendered k-mers present in the unmutated k-mer set no longer exist. The k-mer comparison process is similar to previously described methods for short read error correction, e.g., those described in Greenfield et al., Blue: correcting sequencing errors using consensus and context, Bioinformatics, Volume 30, Issue 19, October 2014, Pages 2723-2732, doi.org / 10.1093 / bioinformatics / btu368.

[0222] minihyb

[0223] In the "minihyb" method, unmutated sequence reads and mutated sequence reads are mapped to a reference sequence. Unmutated sequence reads with mapping positions that overlap with mutated sequence reads trigger gapless realignment of the unmutated sequence reads directly with the mutated sequence reads. This direct unmutated:mutated read alignment can produce better alignments in some cases, particularly when the sequences in the sample DNA deviate from the reference genome sequence. The unmutated:mutated read alignments are then soft-cut using the well-known X-drop method, specifically removing alignments after the peak score position if the alignment score drops more than X from its peak.

[0224] The next step of minihyb uses special methods to select the set of high-scoring alignments for rendering calculations. The strategy applied in this step is to rank the alignments of the unmutated sequence reads according to the alignment score and select the top N alignments in a non-overlapping window of the dream sequence. When the alignment score is formulated using the permutation matrix and gap penalty approach, the score can have a probabilistic interpretation, namely the probability that the unmutated sequence read and the mutated sequence read were generated from the same initial sequence. Selecting the top N alignments effectively imposes a dynamic threshold on the probability, as this threshold can be scaled up or down based on the mutation density in the local region of the dream sequence.

[0225] The selected set of alignments is then used for rendering with the formula defined in Example 4: Adjustment for local sequence composition and unequal mutation rates.

[0226] EM

[0227] In the "EM" method, unmutated sequence reads and mutant sequence reads are first mapped to a reference sequence. As with the minihyb method, unmutated sequence reads with mapping positions that overlap with mutant sequence reads trigger gapless realignment of the unmutated sequence reads directly to the mutant sequence reads. Additional alignments are generated by direct alignment of unmutated sequence reads to mutant sequence reads using a short read aligner capable of multi-mapping unmutated sequence reads onto multiple mutant sequence reads. When rendering a specific mutant sequence read, a set of all unmutated sequence read alignments onto that mutant sequence read is generated, and then expectation-maximization rendering is applied using the EM method 1.

[0228] EM w / cons

[0229] The "EM w / cons" method applies expectation maximization with an additional step of generating a consensus sequence for the unmutated:mutated sequence read alignment. The consensus sequence is generated as described above in the section entitled "Alignment with reference sequence and / or consensus sequence". Once the set of alignments has been collected, the consensus sequence of the mutant sequence reads is discarded and rendered as described in EM method 1.

[0230] Evaluating the Accuracy of Simulated Data

[0231] The simulated data set was analyzed by feeding both mutant and non-mutated sequence reads to the analytical statistical method. This method bypasses the step of assembling mutant sequence reads that would normally be performed in a real application of this technique and has the advantage of eliminating errors in assembling mutant sequence reads from our analysis of the accuracy of the rendering process.

[0232] The accuracy of the simulated data set is evaluated by comparing the rendered long sequence (representing the most likely sequence of the nucleic acid template) with the simulated real unmutated long sequence (representing the real nucleic acid template sequence). This comparison is performed site by site, checking whether each site in the rendered sequence matches its corresponding site in the real sequence.

[0233] When sites in the rendered sequence are identical to the true sequence, these sites are classified as true positives or true negatives. A true positive (TP) is a site that has undergone a mutation and the renderer correctly identifies the mutation and the site is altered to match the original true sequence. A true negative (TN) is a site that has not undergone a simulated mutation and remains unchanged by the rendering process.

[0234] When sites in the rendered sequence differ from the corresponding sites in the true sequence, they are classified as either false positives or false negatives. A false positive (FP) is a site that did not undergo a simulated mutation, but was altered by the rendering process such that it no longer matches a site in the true sequence. A false negative (FN) is a site that underwent a simulated mutation, but the mutation was not correctly identified but was resolved by the rendering.

[0235] Using the above definitions of TP, TN, FP, and FN, various metrics of rendering performance, such as precision and recall, can be calculated. Per-base accuracy scores can be further stratified by QV, for example, FP or FN with very high or very low predicted QVs may be of interest. The results of the accuracy assessment are presented below.

[0236] Results of simulated data

[0237] Three different simulated datasets were generated as described in Table 1. Each dataset was analyzed using the four different rendering methods described above: initial, minihyb, EM, and EM w / cons, except that no results for EM w / cons were present on the medium dataset because the dataset only had 3x coverage of mutant sequence reads, which was insufficient to generate consensus sequences for most of the dataset. The evaluation results for the simulated data are given in Table 2 below.

[0238] Table 2

[0239]

[0240] Several trends emerge from the simulated data. On all data sets, the method using expectation maximization produces significantly better results than the simpler rendering method. Second, the use of consensus sequences leads to improved accuracy on small data sets, but not on large data sets. This difference can be explained by the difference in the way the data sets are processed. Although the small data set uses the same genome sequence to simulate both mutant sequence reads and mapping these mutant sequence reads during consensus sequence generation, the large data set uses the assembly of the HG002 sample for simulation, and the hg38 human genome assembly is used for mutant sequence read mapping and consensus sequence generation. The difference between the hg38 assembly and the HG002 assembly leads to poor consensus sequence generation, which in turn reduces the accuracy of the results.

[0241] Finally, the accuracy for the small and medium datasets is somewhat higher than that for the large dataset. There are several factors that may contribute to this observation. First, the large dataset used real sequence data rather than simulated data for reference reads. Second, as noted above, the large dataset was processed using hg38 as the reference rather than the simulated reads of the HG002 genome. Rendering methods that rely heavily on reference genome mapping (initial, minihyb) appear to be particularly error-prone when there are significant differences between the reference and simulated genomes, resulting in relatively poor performance of those methods.

[0242] Example 10: Accuracy evaluation in real datasets

[0243] The above selection of rendering methods was then applied to a real human genome dataset. The generation of the real human genome dataset was described above. The application of the rendering method was the same as for the simulated data, replacing the simulated non-mutated sequence reads with standard whole genome sequencing (WGS) non-mutated sequencing reads, and replacing the simulated mutant sequence reads with real mutant sequence reads.

[0244] The advantage of evaluating on real datasets is that, to the extent that it is possible to predict the sequence of the initial nucleic acid template with reasonable accuracy, real datasets will capture the effects of sample preparation and sequencing biochemistry that are difficult to simulate.

[0245] A total of 22,945,744 mutant sequence reads were rendered, including 102.2Gbp of sequence data. The N50 of mutant sequence reads in the dataset was 6.2kbp. In the unrendered mutant sequence read data, about 5.4% of the sites showed substitution differences relative to the hg38 reference genome sequence. Some of those differences were true differences between the HG002 sample and the hg38 reference sequence, while most were introduced mutations. In the rendered data, only 0.3% of the sites showed substitution differences relative to the hg38 sample. Therefore, the rendering process greatly reduced the amount of differences relative to the reference.

[0246] Figure 5 Display 500 shows human genome data before and after the rendering process. Figure 5 , three panels are shown in display 500. In panels 510, 520, and 530, each sequence read is shown as a gray horizontal bar, where the sites that are different compared to the hg38 reference sequence are marked with vertical lines. Panel 510 shows the mutant sequence reads of the HG002 sample before rendering. A large number of differences relative to the hg38 reference sequence can be seen in the mutant sequence reads in panel 510. Panel 520 shows the same set of mutant sequence reads compared to the reference sequence after rendering with the minihyb method. As shown in panel 520, after rendering, most of the remaining differences between the rendered sequence and the reference sequence correspond to true variants in the sample, which are shown in panel 530, showing standard WGS unmutated sequence reads from the PCR-free library of the HG002 sample. Therefore, it can be seen that before rendering, in panel 510, the mutant sequence reads contain a large number of differences (mutations) relative to the reference sequence. Rendering removes most of those differences, and the remaining differences are consistent with unmutated sequence reads (reference reads) as true single nucleotide variants (SNVs).

[0247] Other considerations

[0248] It should be understood that all combinations of the foregoing concepts and additional concepts discussed in more detail below (assuming such concepts are not mutually inconsistent) are contemplated as being part of the inventive subject matter disclosed herein. Specifically, all combinations of the claimed subject matter appearing at the end of this disclosure are contemplated as being part of the inventive subject matter disclosed herein. It should also be understood that terms explicitly adopted herein that may also appear in any disclosure incorporated by reference should be given a meaning most consistent with the specific concepts disclosed herein.

[0249] The references to "one example", "another example", "an example", etc. throughout this specification mean that a particular element (such as a feature, structure and / or characteristic) described in conjunction with the example is included in at least one example described herein and may or may not be present in other examples. In addition, it should be understood that the elements used in any example may be combined in various examples in any suitable manner, unless the context clearly indicates otherwise.

[0250] It should be understood that the ranges provided herein include the specified range and any value or sub-range within the specified range, as if such value or sub-range was explicitly listed. For example, a range of about 2 kbp to about 20 kbp should be interpreted as including not only the explicitly listed limits of about 2 kbp to about 20 kbp, but also individual values, such as about 3.5 kbp, about 8 kbp, about 18.2 kbp, etc., and sub-ranges, such as about 5 kbp to about 10 kbp, etc. In addition, when "about" and / or "substantially" are used to describe a value, this means that a small variation (up to + / - 10%) of the stated value is included.

[0251] While several examples have been described in detail, it is to be understood that modifications may be made to the disclosed examples. Therefore, the above description should be considered non-limiting.

[0252] Although certain examples have been described, these examples are presented only as examples and are not intended to limit the scope of the present disclosure. In fact, the novel methods described herein can be embodied in various other forms. Moreover, various omissions, substitutions, and changes can be made in the methods described herein without departing from the spirit of the present disclosure. The attached claims and their equivalents are intended to cover such forms or modifications that will fall within the scope and spirit of the present disclosure.

[0253] Features, materials, characteristics or groups described in conjunction with a particular aspect or example are to be understood to be applicable to any other aspect or example described in this section or elsewhere in this specification unless incompatible therewith. All features disclosed in this specification (including any accompanying claims, abstract and drawings) and / or all steps of any method or process so disclosed may be combined in any combination, except combinations in which at least some of such features and / or steps are mutually exclusive. Protection is not limited to the details of any foregoing examples. Protection extends to any novel feature or any novel combination thereof of the features disclosed in this specification (including any accompanying claims, abstract and drawings), or to any novel step or any novel combination thereof of the steps of any method or process so disclosed.

[0254] Moreover, certain features described in the present disclosure in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments, either individually or in any suitable subcombination. Furthermore, although features may be described above as functioning in certain combinations, in some cases one or more features from the combination may be deleted from a claimed combination, and the combination may be claimed as a subcombination or a variation of a subcombination.

[0255] In addition, although operations may be depicted in the drawings or described in the specification in a particular order, it is not necessary to perform such operations or all operations in the particular order shown or in order to achieve the desired result. Other operations not depicted or described may be incorporated into the example methods and processes. For example, one or more additional operations may be performed before, after, simultaneously or between any described operations. Further, these operations may be rearranged or reordered in other embodiments. Those skilled in the art will appreciate that in some examples, the actual steps taken in the process shown and / or disclosed may be different from the actual steps shown in the figures. Depending on the example, some of the above steps may be removed, or other steps may be added. In addition, the features and attributes of the specific examples disclosed above may be combined in different ways to form additional examples, all of which fall within the scope of the present disclosure.

[0256] For purposes of the present disclosure, certain aspects, advantages, and novel features are described herein. Not all such advantages may be realized according to any particular example. Thus, for example, one skilled in the art will recognize that the present disclosure may be implemented or performed in a manner that achieves one advantage or group of advantages taught herein without necessarily achieving other advantages that may be taught or suggested herein.

[0257] Unless specifically stated otherwise or otherwise understood within the context of use, conditional language such as "may" or "could" is generally intended to convey that certain examples include and other examples do not include certain features, elements, and / or steps. Thus, such conditional language generally does not mean that one or more examples in any way require the features, elements, and / or steps, or that one or more examples must include logic for making decisions with or without user input or prompting, whether or not the features, elements, and / or steps are included in a particular example or are to be performed in any particular example.

[0258] Unless specifically stated otherwise, linking language such as the phrase "at least one of X, Y, and Z" is generally understood in context to convey that an item, term, etc. can be X, Y, or Z. Thus, such linking language is generally not intended to imply that certain examples require the presence of at least one of X, at least one of Y, and at least one of Z.

[0259] As used herein, language of degree, such as the terms "approximately," "about," "generally," and "substantially" represent a value, amount, or characteristic that comes close to the stated value, amount, or characteristic and still performs the intended function or achieves the intended result.

[0260] The scope of the present disclosure is not intended to be limited by the specific disclosure of the examples in this section or elsewhere in this specification, and may be limited by the claims presented in this section or elsewhere in this specification or presented in the future. The language of the claims should be interpreted broadly based on the language employed in the claims, and not limited to the examples described in this specification or during the prosecution of the application, which examples should be understood to be non-exclusive.

Claims

1. A computer-implemented method for determining the sequence of a nucleic acid template by removing mutations present in a mutant sequence read, the method comprising: a) aligning a non-mutated sequence read of the nucleic acid template with a mutated sequence read, wherein the mutated sequence read comprises one or more mutations, and b) Determine: i) the most likely sequence of the nucleic acid template, and ii) for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template; thereby determining the sequence of the nucleic acid template.

2. The method of claim 1, wherein the mutant sequence reads have a mutant nucleic acid template.

3. The method of claim 1 or claim 2, wherein the one or more mutations comprise mutations introduced during sample preparation.

4. The method according to any one of the preceding claims, wherein the one or more mutations comprise mutations introduced randomly during sample preparation by mutagenesis.

5. The method according to any one of the preceding claims, wherein the one or more mutations comprise unintentionally introduced mutations.

6. The method of any one of the preceding claims, wherein the nucleic acid template is longer than the mutated sequence read and the non-mutated sequence read.

7. The method of any one of the preceding claims, wherein i) comprises determining a most likely base for each position associated with an alignment of the non-mutated sequence read with the mutated sequence read.

8. The method according to any one of the preceding claims, wherein i) comprises determining a most likely sequence for a subset of sites associated with the alignment of the non-mutated sequence reads with the mutated sequence reads.

9. The method according to any one of the preceding claims, wherein step b) comprises summarizing the probability of each base at each site associated with the non-mutated sequence read on a site-by-site basis.

10. The method according to any one of the preceding claims, wherein step b) comprises incorporating information relating to the probability of a mutation at each site of the mutated sequence reads.

11. The method according to any of the preceding claims, wherein step b) comprises incorporating information describing the probability of errors associated with the sample preparation or sequencing process.

12. The method according to any one of the preceding claims, wherein step b) comprises determining the most likely base for the site based on the statistically inferred probability of a mutation at the site of a mutant sequence read that is not aligned to a non-mutated sequence read.

13. The method according to any of the preceding claims, wherein step b) comprises incorporating information describing variations in mutation rates between mutant sequence reads or unequal mutation rates of adenine, cytosine, guanosine and thymine.

14. The method according to any one of the preceding claims, wherein the nucleic acid template is derived from a nucleic acid sample comprising two or more copies of a repetitive region of the nucleic acid template.

15. The method of claim 14, wherein step b) comprises estimating the probability that the non-mutated sequence read is derived from a copy of the same repeat region as the mutated sequence read to which the non-mutated sequence read is aligned.

16. A method according to claim 14 or claim 15, wherein step b) comprises applying an optimization process, wherein the optimization process iteratively updates the estimate of the probability that the non-mutated sequence read originates from the same copy as the mutated sequence read to which the non-mutated sequence read is aligned, while updating the inference of the most likely base.

17. The method of claim 16, wherein the optimization process comprises applying an expectation-maximization method.

18. A method according to any one of claims 16 to 17, wherein the optimization process incorporates information describing read length variation.

19. The method of any one of claims 14 to 17, wherein the contribution of each non-mutated sequence read to the probability calculation of the most likely base is weighted by a weighting scheme.

20. The method of claim 19, wherein the weighting scheme is related to the sequence identity of the non-mutated sequence reads and the current iteration estimate of the sequence of the nucleic acid template.

21. The method of claim 14, wherein step b) comprises selecting a subset of non-mutated sequence reads among non-mutated sequence reads derived from different copies of the repeat region.

22. The method of claim 21, wherein selecting a subset of non-mutated sequence reads comprises scoring alignments of non-mutated sequence reads with the mutated sequence reads and selecting the highest scoring alignment.

23. The method of any one of claims 14 to 22, wherein the method comprises determining the sequence of the copies of the repetitive region of the nucleic acid template.

24. The method of any one of claims 14 to 23, wherein the two or more copies of the repeat region comprise two or more haplotypes, and wherein the method comprises determining the sequence of the haplotypes of the nucleic acid sample.

25. The method of any of the preceding claims, wherein the per-base accuracy score comprises a per-base quality value (QV).

26. A method according to any one of the preceding claims, wherein the method further comprises determining an overall accuracy score associated with the most likely sequence.

27. The method of any of the preceding claims, wherein the method further comprises, prior to a), assembling the non-mutated sequence reads, and wherein a) comprises aligning the mutated sequence reads to an assembly of non-mutated sequence reads.

28. The method according to any one of the preceding claims, wherein the method further comprises, prior to a), assembling the mutant sequence reads, and wherein a) comprises aligning the non-mutated sequence reads to the assembly of mutant sequence reads.

29. The method according to any of the preceding claims, wherein the method further comprises, before a), aligning the non-mutated sequence read or the mutated sequence read to a reference sequence.

30. A method according to any of the preceding claims, wherein the method further comprises, before a), assembling the mutant sequence reads, generating a consensus sequence of the assembled mutant sequence reads, and aligning the non-mutated sequence reads with the consensus sequence, and wherein step a) further comprises replacing the consensus sequence with the mutant sequence reads in the alignment of the non-mutated sequence reads with the consensus sequence.

31. The method of claim 28 or claim 30, wherein assembling the mutant sequence reads comprises aligning the mutant sequence reads to a reference sequence.

32. The method of any of the preceding claims, wherein the nucleic acid template comprises a repetitive sequence.

33. The method of claim 4, wherein the randomly introduced mutations comprise substitutions.

34. The method of claim 33, wherein the substitution comprises a transition mutation.

35. The method of any of the preceding claims, further comprising sequencing the unmutated nucleic acid template molecules and the mutated nucleic acid template molecules to obtain unmutated sequence reads and mutated sequence reads.

36. The method of any one of claims 1 to 15, 25 to 26 or 32 to 35, wherein the method comprises: aligning the non-mutated sequence reads and the mutated sequence reads to a reference sequence; determining an initial rendering sequence by determining a most likely base for each position associated with an alignment of the non-mutated sequence read to the mutant sequence read, in combination with information describing variations in mutation rates between mutant sequence reads or unequal mutation rates for adenine, cytosine, guanosine, and thymine; as well as Additional rendered sequences are determined by comparing k-mers from the initial rendered sequence with k-mers from non-mutated sequence reads.

37. The method of any one of claims 1 to 15, 25 to 26 or 32 to 35, wherein the method comprises: aligning the non-mutated sequence reads and the mutated sequence reads to a reference sequence; realigning a non-mutated sequence read having a mapping position that directly overlaps with a mutated sequence read with the mutated sequence read; scoring alignments of non-mutated sequence reads with the mutated sequence reads, and selecting the highest scoring alignment; as well as A rendered sequence is determined by determining a most likely base for each position associated with the alignment of the non-mutated sequence read and the mutated sequence read.

38. The method of any one of claims 1 to 15, 25 to 26 or 32 to 35, wherein the method comprises: The rendering order is determined by: determining a most likely base for each position associated with the alignment of the non-mutated sequence read and the mutated sequence read, and An expectation-maximization (EM) optimization process is applied, wherein the EM optimization process iteratively updates the estimate of the probability that the non-mutated sequence read originates from the same copy as the mutated sequence read to which the non-mutated sequence read is aligned, while simultaneously updating the inference of the most likely base.

39. The method of any one of claims 1 to 15, 25 to 26 or 32 to 35, wherein the method comprises: generating a consensus sequence of the assembled mutant sequence reads; aligning the unmutated sequence reads to the consensus sequence; in said alignment of said non-mutated sequence reads with said consensus sequence, replacing said consensus sequence with said mutated sequence read; and The rendering order is determined by: determining a most likely base for each position associated with the alignment of the non-mutated sequence read and the mutated sequence read, and An expectation-maximization (EM) optimization process is applied, wherein the EM optimization process iteratively updates the estimate of the probability that the non-mutated sequence read originates from the same copy as the mutated sequence read to which the non-mutated sequence read is aligned, while simultaneously updating the inference of the most likely base.

40. A system comprising a processor configured to perform a method comprising: a) aligning a non-mutated sequence read of the nucleic acid template with a mutated sequence read of the mutated nucleic acid template, wherein the mutated sequence read comprises one or more mutations, and b) Determine: i) the most likely sequence of the nucleic acid template, and ii) for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template; thereby determining the sequence of the nucleic acid template.

41. A computer readable medium comprising instructions which, when executed by a processor, perform a method comprising: a) aligning a non-mutated sequence read of the nucleic acid template with a mutated sequence read of the mutated nucleic acid template, wherein the mutated sequence read comprises one or more mutations, and b) Determine: i) the most likely sequence of the nucleic acid template, and ii) for each base of the most likely sequence, a per-base accuracy score associated with the probability that the base matches the nucleic acid template; thereby determining the sequence of the nucleic acid template.

Images