Method for dynamically setting short sequence variation blank limit

Abstract

The invention establishes a method for dynamically setting a short sequence variation blank limit, and strictly utilizes the information of a blank sample to specifically and dynamically define the variation LoB for the variation depth to be measured. According to the method, LoBs of all SNVs and indel in the ROI of the detected sample can be defined, and the accuracy of variation detection is improved.

Description

Technical Field

[0001] This invention relates to the field of variant detection technology. More specifically, this invention relates to a method for dynamically setting short sequence variant blank limits. Background Technology

[0002] Next-generation sequencing (NGS) can detect a variety of variations in the genome sequence, including short sequence variants (SSVs) and structural variants (SVs). SSVs are the most common and the most mature variants to detect, and they are widely used in genetic disease screening and diagnosis, tumor typing, medication guidance, early diagnosis and screening, and recurrence detection, and have important clinical value.

[0003] The LoB (LoB) is a crucial parameter of a detection system, reflecting its background noise level and signal detection capability. The definition and estimation of LoB is a classic problem in detection science, with general methods discussed in numerous standards, such as CLSI EP17-A2 (Evaluation of Detection Capability for Clinical Laboratory Measurement Procedures; Approved Guideline—Second Edition). However, the LoB for SSVs in NGS has its unique characteristics. First, as a high-throughput method, NGS can simultaneously detect tens of thousands of variants, each with a potentially different LoB. Second, because the amount of data from tested samples often fluctuates within a range, and the depth at specific variant locations is inconsistent, a fixed LoB usually leads to performance degradation. The LoB for SSVs is generally expressed as variant allele frequency (VAF). If the LoB is set too high, more false negatives are produced, and vice versa. Therefore, a more accurate definition and estimation of the LoB is needed.

[0004] There are two main categories of publicly disclosed methods for setting LoB: (1) Subjectively setting a fixed LoB that covers the entire ROI or specific variant groups (e.g., setting hotspot and non-hotspot variants separately) within the target region (ROI), such as the methods disclosed in patents CN107944223B and CN108690871B; (2) For different variants within the ROI, based on the mean of the blank sample VAF. blank and standard deviation SD blank And the z-score corresponding to the statistical confidence level, calculate LoB = Mean blank +z×SD blank Such as the method disclosed in patent US20190073445A1.

[0005] However, these techniques all have drawbacks. First, they do not define a LoB (Lower Opposite B) on a per-variable basis. The human genome is not homogeneous; genomic location and sequence variations together define a variant. The noise of a variant is affected by the complexity of its upstream and downstream sequences in the genome, as well as by the chemical properties of the base itself. Therefore, different variants have different noise backgrounds in the detection system. Setting a generalized LoB makes it difficult to precisely control detection performance. This is the drawback of the first type of method mentioned above.

[0006] Second, the method for setting the LoB is not rigorous. Artificially setting the LoB by observing the signal of blank samples cannot avoid the influence of the observer's subjectivity, does not follow statistical analysis methods, and is difficult to provide a reasonable interpretation of the test results. This is a drawback of the first type of method mentioned above. Estimating the LoB using standard deviation and z-scores assumes that the noise VAF follows a normal distribution. However, the noise VAFs obtained from different blank samples in different sequencing batches generally do not follow a normal distribution. Therefore, using a normal distribution to set the LoB is inaccurate. This is a drawback of the second type of method mentioned above.

[0007] Third, the LoB is set statically, rather than dynamically. Both of the above methods implicitly assume that sequencing depth is stable across different genomic regions, failing to consider the impact of varying sequencing depths on the LoB. When the depth of the target site becomes abnormal in a particular assay, the LoB, not calculated based on depth information, may become invalid.

[0008] Therefore, there has always been a need in the art for a reliable and efficient method for setting the LoB of SSV. Summary of the Invention

[0009] This invention establishes a method for setting the LoB of SNVs, which strictly utilizes information from blank samples to dynamically define the LoB of variants specifically for the depth of the variant to be detected, for sample detection in UMI-independent NGS. This method can define the LoB of all SNVs and indels within the ROI of the tested sample, improving the accuracy of variant detection.

[0010] In one aspect, the present invention provides a method for detecting short sequence variants in a sample, the method comprising: estimating background noise of one or more short sequence variants using PoN, obtaining the LoB of the short sequence variant based on the background noise and the sequencing depth of the sample to be tested, and determining the detection authenticity of the short sequence variant in the sample to be tested based on the LoB.

[0011] In one embodiment, the method further includes obtaining sequencing data of normal samples in the PoN, the sequencing data covering the genomic locations of short sequence variations.

[0012] In one embodiment, the method further includes distinguishing germline variation from noise variation based on a genotype vector including a reference sequence and a genotype depth vector at genomic locations covering short sequence variations.

[0013] In one embodiment, the method further includes determining the VAF (Variation Aspect-Oriented Vector) for any short sequence variant when it appears as a noisy variant in the PoN sample, thereby constructing a background noise vector for the short sequence variant. In one embodiment, the method further includes constructing a background noise vector for the short sequence variant for PoN in an NGS detection scenario without using UMI.

[0014] In one embodiment, the method further includes determining the sequencing depth of the sample at short sequence variant locations and a preset specificity.

[0015] In one embodiment, the method further includes determining short sequence variations based on the VAF of short sequence variations in the test sample.

[0016] In another aspect, the present invention provides a method for detecting short sequence variations in a sample by dynamically defining the variation LoB in units of short sequence variations, the method comprising:

[0017] - Obtain sequencing data from normal samples in the PoN, the sequencing data covering the genomic locations of short sequence variations;

[0018] - Identify germline and noise variations detected in PoN samples at the genomic locations based on genotype vectors including reference sequences and genotype depth vectors;

[0019] - Determine the VAF when any short sequence variant appears as a noise variant in the PoN sample;

[0020] -For PoN, construct a background noise vector for short sequence mutations;

[0021] - Determine the sequencing depth and predefined specificity at short sequence variant sites in the sample to be tested;

[0022] - Calculate LoB based on the sequencing depth and specificity, as well as the background noise vector of short sequence variants; and

[0023] - The detection validity of short sequence variants is determined based on LoB. When the VAF of the short sequence variant in the test sample is greater than or equal to LoB, the short sequence variant is considered positive; when the VAF of the short sequence variant in the test sample is less than LoB, the short sequence variant is considered negative.

[0024] In another aspect, the present invention provides a method for obtaining a background noise vector of short sequence mutations from PoN, the method comprising:

[0025] - For a specific genomic location in a PoN sample, obtain all genotypes at that location, including the reference sequence, and sort them in descending order of depth to obtain a genotype vector and a vector corresponding to the depth of each genotype;

[0026] -Based on the vector of genotype depth, germline variation and noise variation are distinguished by the cumulative distribution function of the binomial distribution;

[0027] - For any variant, obtain the VAF when it is detected as a noise variant in a sample of PoN; and

[0028] -Based on maximum likelihood estimation, the background noise vector of short sequence variations for PoN is obtained.

[0029] In another aspect, the present invention provides a method for obtaining LoBs of short sequence variants from PoN, the method comprising:

[0030] - For a specific genomic location in a PoN sample, obtain all genotypes at that location, including the reference sequence, and sort them in descending order of depth to obtain a genotype vector and a vector corresponding to the depth of each genotype;

[0031] -Based on the vector of genotype depth, germline variation and noise variation are distinguished by the cumulative distribution function of the binomial distribution;

[0032] - For any variant, obtain the VAF when it is detected as a noise variant in a sample of PoN;

[0033] -Based on maximum likelihood estimation, the background noise vector for short-sequence variations of PoN is obtained; and

[0034] -Calculate the LoB of the short sequence variant based on the sequencing depth and preset specificity of the sample at the short sequence variant location, as well as the background noise vector.

[0035] In some implementations of various aspects, the short sequence variants may include single nucleotide variants (SNVs) and insertion / deletion variants (indels).

[0036] In some implementation schemes, the sample to be tested is a normal sample, a suspected lesion sample, or a lesion sample.

[0037] In some implementations of various aspects, all samples in the PoN are normal samples.

[0038] In some implementations of various aspects, the normal sample includes tissue from healthy humans or healthy tissue from diseased subjects, such as healthy tissue from cancer patients, i.e., adjacent tissue.

[0039] In one embodiment, the method of the present invention is executed by a computer program. The computer program may include one or more modules for performing one or more steps of the method of the present invention. The computer program may be stored on a memory or medium, such as magnetic tape, magnetic disk, optical disk, ROM, PROM, VCD, DVD, or other computer-readable media. Those skilled in the art will understand that the computer program itself is a conventional technique in the field. How to obtain and execute the computer program after reading this disclosure is entirely within the capabilities of those skilled in the art.

[0040] Therefore, in another aspect, the present invention provides a computer device including a memory, a processor, and a computer program stored on the memory, the processor executing the computer program to implement the method of the present invention. In yet another aspect, the present invention provides a computer-readable storage medium storing a computer program thereon, the computer program implementing the method of the present invention when executed by a processor. In still another aspect, the present invention provides a computer program product including a computer program that implements the method of the present invention when executed by a processor. Attached Figure Description

[0041] Figure 1 : A schematic flowchart of the present invention.

[0042] Figure 2 In the method example of the present invention, the mutation direction distribution of the SNV of the test sample is shown.

[0043] Figure 3 In the method example of the present invention, the frequency of occurrence of indel in the test sample. Detailed Implementation

[0044] To fully understand the purpose, features, and effects of this invention, the following specific embodiments will be used to describe the invention in detail. Unless otherwise stated, the technical terms used in this invention have the meanings commonly understood by those skilled in the art.

[0045] I. Definition

[0046] High-throughput sequencing (HTS), also known as massively parallel sequencing (MPS) or next-generation sequencing (NGS), is a sequencing technology characterized by its ability to sequence hundreds of thousands to billions of nucleic acid molecules in parallel and its short read lengths.

[0047] Unique molecular identifier (UMI): A short nucleotide sequence used to uniquely label the original nucleic acid molecule in order to correct base sequence errors introduced in library construction, sequencing and other steps in the results.

[0048] Short sequence variation (SSV): Short differences from the reference genome sequence found in the tested sample, including single-nucleotide variation (SNV) and insertion / deletion (indel) variations not exceeding 40 bp in length.

[0049] Structural variation (SV): Large changes in the structure of the genome relative to the reference genome found in the tested sample, which are generally considered to include at least copy number variation (CNV) and gene fusion.

[0050] Limit of blank (LoB): The "highest" measurement result that can be observed in a blank sample within a given detection system. The probability of detecting a blank sample is higher than LoB, expressed as α.

[0051] Regions of interest (ROI): also known as "tested area / region", "region of interest / region", "target area / region", etc., are the target genomic regions that are preset in the genetic testing product and that are of interest to developers and users.

[0052] Variant allele frequency (VAF): The proportion of the total coverage depth of a genome locus that supports a variant.

[0053] Blank samples, also known as control samples in some scenarios, are negative samples determined using orthogonal information. They are considered to not carry the detection target. Target signals detected in control samples are considered noise.

[0054] Normal sample panel (PoN): This is the set of normal samples (as opposed to lesion samples). All samples in the PoN are considered blank samples.

[0055] Formalin fixation and paraffin embedding (FFPE): To protect the integrity of cell structure, tissue samples are first fixed with formalin and then embedded in paraffin. Tissue samples that are easily cut into thin slices are called FFPE samples. They can be preserved at room temperature for a relatively long time or used to prepare tissue specimens for testing.

[0056] II. The Method of the Invention

[0057] The following is a detailed description of the solution of this invention, which is named WeaPoN ( We ighted-CDF b a sedon P anel o f N (ormals). Schematic flowchart as follows: Figure 1 As shown, PoN is used here to estimate background noise, thereby determining LoB. PoN consists of normal samples, which can be tissue from healthy individuals or healthy tissue (adjacent tissue) from cancer patients.

[0058] 1. Use PoN to obtain the background noise vector of the mutated v.

[0059] a) Differentiate between germline variation and noise variation

[0060] For a specific genomic location in a PoN sample, obtain all genotypes at that location, including the reference sequence, and sort them in descending order of depth to obtain a genotype vector:

[0061] (a1, a2, ..., a) n )

[0062] And the vector corresponding to the depth of each genotype:

[0063] (d1, d2, ..., d n ), where d1≥d2≥…≥d n

[0064] make:

[0065] D = sum(d1, d2, ..., d...) n )

[0066] Based on the fact that humans are diploid organisms:

[0067] 1) If d1 > d2 + d3 + ... + d n And pbinom(D-d1, D, p) < P cutoff If the locus is homozygous, then a1 is a germline variation, and the rest are noise variations.

[0068] 2) If d1 > d2 + d3 + ... + d n And pbinom(D-d1, D, p)≥P cutoff If the locus is heterozygous, then a1 and a2 are germline variations, and the rest are noise variations.

[0069] 3) If d1≤d2+d3+…+d n If the locus is heterozygous, then a1 and a2 are germline variations, and the rest are noise variations.

[0070] Where pbinom(k, n, p) is the cumulative distribution function (CDF) of the binomial distribution; p is the gene frequency of heterozygous variation, typically taken as 0.5; P cutoff Users can set the parameters according to their expectations of the probability of statistical errors.

[0071] b) Obtain the background noise vector

[0072] For any variant v, the VAF (denoted as NVAF) when it is detected as a noise variant in a sample i of the PoN can be obtained using the above method, that is:

[0073]

[0074] In NGS detection scenarios without UMI, it can be assumed that the error rate r follows a binomial distribution, leading to background noise (see Jennings, Lawrence J., Maria E. Arcila, Christopher Corless, Suzanne Kamel-Reid, Ira M. Lubin, John Pfeifer, Robyn L. Temple-Smolkin, Karl V. Voelkerding, and Marina N. Nikiforova. "Guidelines for Validation of Next-Generation Sequencing-Based Oncology Panels: A Joint Consensus Recommendation of the Association for Molecular Pathology and College of American Pathologists." The Journal of Molecular Diagnostics: JMD 19, no. 3 (May 2017): 341-65. https: / / doi.org / 10.1016 / j.jmoldx.2017.01.011.). Based on the idea of ​​maximum likelihood estimation, we have: r vi =NVAF vi .

[0075] Therefore, for a PoN containing M blank samples, we can obtain the background noise vector of the mutated v:

[0076] V v =(r v1 ,...,r vi ,...,r vM )=(NVAF v1 , ..., NVAF vi , ..., NVAF vM )

[0077] 2. Calculate the LoB of the variation v based on the background noise vector and the depth of the test sample.

[0078] a) Let the sequencing depth of the sample at the mutation v position be d. Given a preset specificity of Sp, the minimum q value that makes the following formula true is the LoB of the sample for mutation v in the current test:

[0079]

[0080] In one specific implementation, the specificity Sp is 95%, 99%, 99.9%, or 99.99%.

[0081] b) The detection validity of short sequence variants in the test sample is determined based on LoB. When the VAF of the short sequence variant in the test sample is greater than or equal to LoB, the short sequence variant is determined to be positive; when the VAF of the short sequence variant in the test sample is less than LoB, the short sequence variant is determined to be negative.

[0082] Compared with the prior art, the advantages of the present invention are as follows:

[0083] 1. LoB definition is more granular. WeaPoN can define LoBs for all possible SSVs within an ROI, providing a unique LoB for each possible SSV. Compared to LoBs for a specific genomic segment or a certain type of variant, WeaPoN fully adapts to the heterogeneity of variants.

[0084] 2. The definition process of LoB takes into account a complete range of dimensions. This is reflected in: (1) setting a specific tolerance for Type I errors, rather than subjectively and arbitrarily defining it by observing historical case data or by statistical assumptions that do not conform to reality; (2) incorporating depth factors, so that the calculation of LoB is more in line with the characteristics of high-throughput sequencing data and can be dynamically adjusted according to actual sequencing data.

[0085] The present invention is further illustrated below by way of embodiments, but the invention is not limited to the scope of the embodiments described herein. Experimental methods in the following embodiments that do not specify specific conditions were performed according to conventional methods and conditions, or as selected according to the product instructions.

[0086] Example 1

[0087] The sample used in this embodiment is 3000 tumor samples prepared by FFPE. The specific steps for SSV detection using WeaPoN are as follows:

[0088] 1. Sequencing data from 1030 adjacent normal samples prepared by FFPE were used as PoN, with a total detection interval length of 2,129,706 bp. All genomic loci with potential SNV and indel variants were traversed within the detection range.

[0089] 2. For a specific variant in a given adjacent normal sample, obtain the depth of all genotypes at the location of that variant and sort them in descending order. Substitute the above data into d1, d2+d3+…+d i The algorithm `pbinom(D-d1, D, p)` is used to determine whether each genotype is a noise variant. If a variant genotype is defined as a noise variant, its error rate is the VAF; otherwise, it is 0.

[0090] 3. For a certain background noise vector (r) v1 ,...,r vi ,...,r v1030 Given the variation v of the tumor sample to be tested, and assuming a specificity of 99.99%, find the minimum value of q that makes the following formula hold true. This value is the LoB of the variation v:

[0091]

[0092] 4. A positive result is indicated if the VAF of the variant v in the sample is greater than or equal to LoB; otherwise, it is considered negative.

[0093] Using a fixed LoB strategy (variables with VAF ≥ 0.02 are considered positive), the results of SNV and indel comparisons are shown in Tables 1 and 2, respectively. For SNV, among the variants that were negative by WeaPoN but positive by fixed LoB, the C > T type was dominant. Figure 2 The vertical axis represents the number of reference SNVs, consistent with the type of incorrect bases introduced by FFPE treatment (see Do, H. & Dobrovic, A. Sequence artifacts in DNA from formalin-fixed tissues: causes and strategies for minimization. Clin. Chem. 61, 64–71 (2015)). Furthermore, compared to variants identified as positive by WeaPoN but negative by fixed LoB, only a few variants in this subset contributed significantly more to the latter's observed number in tumor samples (Table 1 and...). Figure 2 This indicates that these variants are frequently repeated, further suggesting the possibility that they are noise. For indel, variants that were negative by WeaPoN but positive by fixed LoB were contributed by 678 variants, and they were frequently repeated. Figure 3 The vertical axis represents the number of observations for each indel, suggesting that these variants may be systemic noise; variants that are positive by WeaPoN but negative by fixed LoB have a highest frequency of only 30 (where the vertical axis represents the number of observations for each indel), indicating ... Figure 3 This represents approximately 1% of the total sample, suggesting that these mutations may be real. The above data indicates that the WeaPoN algorithm is more effective at controlling noise than the strategy of fixing the LoB.

[0094] Table 1: Comparison of SNV results between WeaPoN and Fixed LoB strategies

[0095]

[0096] Table 2 Comparison of indel results between WeaPoN and Fixed LoB strategies

[0097]

[0098]

[0099] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any substitutions, modifications, combinations, changes, simplifications, etc., made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for detecting short sequence variants in a sample, characterized in that the method comprises: - estimating background noise of one or more short sequence variants using PoN; - obtaining LoB of the short sequence variants based on the background noise and sequencing depth of the sample to be tested; and - determining the detection authenticity of the short sequence variants in the sample to be tested according to the LoB. 2.The method of claim 1, characterized in that the method further comprises obtaining sequencing data of normal samples in the PoN, which covers the genomic positions of the short sequence variants. 3.The method of claim 1, characterized in that the method further comprises distinguishing germline variants from noise variants according to genotype vectors including reference sequences and genotype depth vectors at the genomic positions covering the short sequence variants. 4.The method of claim 1, characterized in that the method further comprises determining VAF of any short sequence variant as a noise variant in the PoN samples, thereby constructing a background noise vector of the short sequence variants. 5.The method of claim 4, characterized in that the method further comprises constructing the background noise vector of the short sequence variants for the PoN in the NGS detection scenario without using UMIs. 6.The method of claim 1, characterized in that the method further comprises determining the sequencing depth of the sample to be tested at the position of the short sequence variant and the preset specificity. 7.The method of claim 1, characterized in that the method further comprises making a decision on the short sequence variant based on the VAF of the short sequence variant in the sample to be tested. 8.A method for detecting short sequence variants in a sample by dynamically defining variant LoB in units of short sequence variants, characterized in that the method comprises: - obtaining sequencing data of normal samples in the PoN, which covers the genomic positions of the short sequence variants; - identifying germline variants and noise variants detected by the PoN samples at the genomic positions according to genotype vectors including reference sequences and genotype depth vectors; - determining VAF of any short sequence variant as a noise variant in the PoN samples; - constructing a background noise vector of the short sequence variants for the PoN; - determining the sequencing depth of the sample to be tested at the position of the short sequence variant and the preset specificity; - calculating LoB based on the sequencing depth and specificity and the background noise vector of the short sequence variants; and - determining the detection authenticity of the short sequence variants according to the LoB, wherein when the VAF of the short sequence variant in the sample to be tested is greater than or equal to the LoB, the short sequence variant is determined to be positive; and when the VAF of the short sequence variant in the sample to be tested is less than the LoB, the short sequence variant is determined to be negative. 9.A method for obtaining a background noise vector of short sequence variants from a PoN, characterized in that the method comprises: - obtaining all genotypes including reference sequences at a certain genomic position of a certain PoN sample and arranging them in descending order of their depths to obtain a genotype vector and a vector corresponding to the depth of each genotype for a certain genomic position of a certain PoN sample. ​ ​ - according to the vector of genotype depth, distinguishing germline variants from noise variants by the cumulative distribution function of binomial distribution; - for any variant, obtaining the VAF when it is detected as a noise variant in a sample of PoN; and - obtaining the background noise vector of short sequence variants for PoN according to maximum likelihood estimation.

10. A method for obtaining LoB of short sequence variants from PoN, characterized in that the method comprises: - obtaining all genotypes of a genomic position of a sample of PoN, including the reference sequence, and arranging them in descending order of depth, obtaining a vector of genotypes and a vector corresponding to the depth of each genotype; - according to the vector of genotype depth, distinguishing germline variants from noise variants by the cumulative distribution function of binomial distribution; - for any variant, obtaining the VAF when it is detected as a noise variant in a sample of PoN; - obtaining the background noise vector of short sequence variants for PoN according to maximum likelihood estimation; and - calculating the LoB of the short sequence variant according to the sequencing depth of the sample to be tested at the short sequence variant position, the preset specificity, and the background noise vector.

11. The method of any one of claims 1-10, characterized in that the short sequence variant can include single nucleotide variants (SNVs) and insertion / deletion variants (indels).

12. The method of any one of claims 1-8, characterized in that the sample to be tested is a normal tissue sample, a suspected lesion sample, or a lesion sample.

13. The method of any one of claims 1-10, characterized in that all samples in the PoN are normal tissue samples.

14. The method of claim 13, characterized in that the normal tissue sample includes tissue of a healthy human or healthy tissue of a diseased subject.

15. The method of claim 14, characterized in that the normal tissue sample is healthy tissue of a tumor patient.

16. A computer apparatus comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program, when executed by the processor, causes the processor to perform the method of any one of claims 1 to 15. The processor executes the computer program to implement the method of any one of claims 1-15.

17. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the method of any one of claims 1-15.

18. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1-15.

Images