Apparatus and method for detecting snvs

Abstract

The present disclosure relates to methods, devices, and systems for detecting SNVs of a target gene in a sample with high accuracy.

Description

Technical Field

[0001] This invention belongs to the field of gene detection technology, and in particular relates to a method, system and apparatus for detecting SNVs of target genes in samples with high accuracy. Background Technology

[0002] Genetic mutations are key factors influencing tumor development and progression. Next-generation sequencing (NGS) is currently the mainstream and most promising technique for detecting the location and type of tumor cell mutations. Among the various types of mutations, single nucleotide variations (SNVs) are the most common. The NGS SNV analysis workflow consists of the following steps:

[0003] 1) Remove low-quality sequences and foreign sequences (such as sequencing adapters and molecular tags) from the processed data.

[0004] 2) The sequences were aligned back to the reference genome using sequence alignment software, and duplicate sequences from the same DNA template were removed to obtain the deduplicated BAM file.

[0005] 3) Compare the signal carrying the variant site in the BAM file with the background noise. When the signal of the variant site is significantly higher than the background noise, the mutation at that site is considered positive.

[0006] In step 3) above, the accuracy of the background noise model and its estimation directly affects the accuracy of the results. NGS data includes errors caused by various experimental factors (collectively referred to as background noise), affecting the accuracy of data analysis. Therefore, determining tumor SNV requires an accurate estimate of the background noise in the experiment. Currently, the common estimation method involves collecting a certain number of normal human samples (panel of normal, PON), treating the signals in the PON as background noise, and establishing a statistical model for the background noise.

[0007] The existing LoLoPicker scheme uses normal human samples as PON, filters out possible real variations through the K-means clustering algorithm, and then uses a binomial model to estimate background noise.

[0008] Under the existing technical framework, the PON method has the following main drawbacks: 1) The number of PON samples is usually very limited, making it impossible to accurately assess background noise; 2) PON samples are usually selected from healthy individuals, while clinical samples from hospital oncology departments are usually from cancer patients or high-risk individuals, whose background noise differs in some genes; 3) Once PON is selected, the background noise model is also fixed, and the testing service provider cannot update it in a timely manner.

[0009] The purpose of this invention is to solve the three problems mentioned above. Background noise estimation will no longer use PON data from healthy individuals, but rather clinical samples of cancer patients accumulated by the testing service provider, thereby obtaining a more accurate noise estimation model. Summary of the Invention

[0010] The inventors of this application have pioneered the use of clinical samples instead of healthy human samples to establish a background noise model; and proposed a novel truncated statistical model to ensure the authenticity of negative samples used for model fitting, thereby obtaining a more accurate background noise estimate; at the same time, they have gradually optimized the noise estimate using continuously accumulated clinical samples.

[0011] Based on this, the present invention provides a method, system and apparatus for detecting SNVs of target genes in samples with high accuracy, ensuring that only highly certain negative samples are used to estimate noise, thereby obtaining more accurate background noise estimates.

[0012] The method of the present invention mainly includes the following four steps: 1) extracting site information from BAM files; 2) determining the threshold for negative samples; 3) establishing a statistical model for negative samples and estimating parameters; and 4) determining whether there is variation at the site in the tested sample.

[0013] Compared with the prior art, the present invention has at least the following advantages:

[0014] 1) Compared to PON composed of normal human samples, the clinical samples used in this invention are more representative of the complexity of tumor samples;

[0015] 2) Existing clustering methods cannot completely remove true variations, and may contain low-frequency variations. The threshold selection method established in this invention ensures that only highly specific negative samples are used to estimate noise;

[0016] 3) The truncated beta binomial model established in this invention is more flexible and more suitable for noise estimation across samples;

[0017] 4) According to the present invention, the testing provider can update the background noise assessment model in a timely manner based on the accumulation of clinical samples.

[0018] In the first aspect, this disclosure provides a method for detecting samples with high accuracy. i A device for detecting the SNV of a target gene, the device comprising: a sequencing data reading module for performing step (A) quality control on the raw sequencing data, aligning it to a reference genome and removing duplicates, and generating a BAM file that stores sequencing fragment (reads) information;

[0019] The VAF calculation module is used to perform step (B) extracting the number of reads carrying variations at base sites on the genomic sequence from the deduplicated BAM file (x). i ) and sequencing depth (n i ), through x i / n i Calculate the variation frequency (VAF);

[0020] The threshold determination module is used to implement step (C) to calculate the threshold c using the Gaussian smoothing kernel function of the VAF distribution of the site, and classify samples that do not exceed c as negative;

[0021] The background noise estimation module is used to implement step (D) of selecting negative samples. i and n i The background noise parameter θ is estimated by fitting a truncated beta-binomial distribution and using the maximum likelihood method.

[0022] The SNV determination module is used to implement step (E) to determine the number of variant reads (x). i A threshold is set, and samples exceeding the threshold are identified as variants.

[0023] In one implementation, only when x i / n i When c ≤ c, there is no true variation in the samples used to estimate background noise.

[0024] In one implementation, in step (C), the threshold c is set as follows: (The remaining text appears to be incomplete and requires further context.) t i = log ( x i / n i ),right t i Apply Gaussian kernel smoothing to the distribution:

[0025]

[0026] in m For the sample size, s 2 The variance of the Gaussian kernel function. exp () is an exponential function. K ( t () is the kernel smoothed t i The distribution function;

[0027] right K ( t Find the maximum value and local minimum value respectively:

[0028]

[0029] Select VAF < t l For the samples, set a threshold c This is the 80th percentile of the VAF for this portion of the sample.

[0030] In one implementation, step (D) is performed as follows: for all negative samples (including VAF ≤ 0.5%) c and VAF > c )of x i and n i Establish a statistical model:

[0031]

[0032] in Let be the probability mass function of the beta-binomial. For the two-dimensional parameters of the model, For beta functions, ! represents the gamma function, and ! represents the factorial sign.

[0033]

[0034] Other notes The cumulative distribution function of the beta binomial:

[0035]

[0036] Exclude step (C) x i / n i > c After clinical samples, the number of variations The statistical distribution becomes a right-truncated beta binomial distribution:

[0037]

[0038] Estimating background noise parameters using the maximum likelihood method : Let be the log-likelihood function, when hour, Take the maximum value. This is the maximum likelihood estimate of the background noise, and Determine the distribution of background noise.

[0039] .

[0040] In one implementation, step (E) is performed as follows: for the tested sample ( x , n ),when p The value is less than the set false positive rate threshold. α The sample was determined to be positive at that time.

[0041] .

[0042] In one embodiment, the sample is a tumor sample, and the tumor is selected from solid tumors.

[0043] In one embodiment, the pan-cancer solid tumor is selected from ovarian cancer, breast cancer, colorectal cancer, lung cancer, prostate cancer, liver cancer, stomach cancer, pancreatic cancer, cervical cancer, bladder cancer, kidney cancer, melanoma, and head and neck cancer, etc.

[0044] In one implementation, the target gene is any gene contained in the original sequencing data.

[0045] In one embodiment, the target gene is the BRAF gene, and the site is the V600E site.

[0046] Secondly, this disclosure provides a method for detecting samples with high accuracy. i A method for SNV of a target gene in the middle, the method comprising:

[0047] Sequencing data reading step (A): Perform quality control on the raw sequencing data, align it to the reference genome and remove duplicates, and generate a BAM file that stores sequencing fragment (read) information;

[0048] VAF calculation step (B): Extract the number of reads carrying variations from the base sites on the genomic sequence (x) from the deduplicated BAM file. i ) and sequencing depth (n i ), through x i / n i Calculate the variation frequency (VAF);

[0049] Threshold determination step (C): Calculate the threshold c using the Gaussian smoothing kernel function of the VAF distribution at this site, and classify samples that do not exceed c as negative;

[0050] Background noise estimation step (D): Select x of the negative sample i and n i The background noise parameter θ is estimated by fitting a truncated beta-binomial distribution and using the maximum likelihood method.

[0051] SNV determination step (E): Determine the number of mutated reads (x iA threshold is set, and samples exceeding the threshold are identified as variants.

[0052] In one implementation, only when x i / n i When c ≤ c, there is no true variation in the samples used to estimate background noise.

[0053] In one implementation, in step (C), the threshold c is set as follows: (The remaining text appears to be incomplete and requires further context.) t i = log ( x i / n i ),right t i Apply Gaussian kernel smoothing to the distribution:

[0054]

[0055] in m For the sample size, s 2 The variance of the Gaussian kernel function. exp () is an exponential function. K ( t () is the kernel smoothed t i The distribution function;

[0056] right K ( t Find the maximum value and local minimum value respectively:

[0057]

[0058] Select VAF < t l For the samples, set a threshold c This is the 80th percentile of the VAF for this portion of the sample.

[0059] In one implementation, step (D) is performed as follows: for all negative samples (including VAF ≤ 0.5%) c and VAF > c )of x i and n i Establish a statistical model:

[0060]

[0061] in Let be the probability mass function of the beta-binomial. For the two-dimensional parameters of the model, For beta functions, ! represents the gamma function, and ! represents the factorial sign.

[0062]

[0063] Other notes The cumulative distribution function of the beta binomial:

[0064]

[0065] Exclude step (C) x i / n i > c After clinical samples, the number of variations The statistical distribution becomes a right-truncated beta binomial distribution:

[0066]

[0067] Estimating background noise parameters using the maximum likelihood method : Let be the log-likelihood function, when hour, Take the maximum value. This is the maximum likelihood estimate of the background noise, and Determine the distribution of background noise.

[0068]

[0069] .

[0070] In one implementation, step (E) is performed as follows: for the tested sample ( x , n ),when p The value is less than the set false positive rate threshold. α The sample was determined to be positive at that time.

[0071] .

[0072] In one embodiment, the sample is a tumor sample, and the tumor is selected from solid tumors.

[0073] In one embodiment, the pan-cancer solid tumor is selected from ovarian cancer, breast cancer, colorectal cancer, lung cancer, prostate cancer, liver cancer, stomach cancer, pancreatic cancer, cervical cancer, bladder cancer, kidney cancer, melanoma, and head and neck cancer, etc.

[0074] In one embodiment, the target gene is the BRAF gene, and the site is the V600E site.

[0075] In one implementation, the target gene is any gene contained in the original sequencing data.

[0076] In a third aspect, this disclosure provides a computer system for implementing any of the methods described in the second aspect, comprising:

[0077] Input device, used to input sequencing information of the sample's nucleic acid sequence;

[0078] Computer memory is used to store computer program instructions;

[0079] A computer processor is configured to execute the computer program instructions, wherein the computer program instructions perform the SNV detection and transmit the results to an output device; and

[0080] An output device for displaying the results.

[0081] In a fourth aspect, this disclosure provides a computer-readable medium in which...

[0082] The computer-readable medium stores a computer program.

[0083] The computer program described herein can be executed by a computer processor to implement any of the methods described in the second aspect.

[0084] The following description and examples illustrate embodiments of the present invention in detail. It should be understood that the present invention is not limited to the specific embodiments described herein and therefore can be modified. Those skilled in the art will recognize that many variations and modifications exist in the present invention, all of which are included within its scope. Attached Figure Description

[0085] Figure 1 The flowchart of one embodiment of the present disclosure includes: 1) extracting site information from the BAM file; 2) determining the threshold for negative samples; 3) establishing a statistical model for negative samples and estimating parameters; and 4) determining whether there is variation at the site in the tested sample. Detailed Implementation

[0086] The inventors of this invention have pioneered the use of clinical samples, rather than healthy human samples, to establish a background noise model and propose a novel truncated statistical model. Furthermore, the background noise model and truncated statistical model ensure the authenticity of the negative samples used for model fitting, thereby obtaining a more accurate background noise estimate. Simultaneously, the noise estimate can be progressively optimized using continuously accumulating clinical samples.

[0087] As mentioned above, existing technologies use normal human samples as PONs, which cannot completely remove real variations and therefore cannot accurately assess background noise. This disclosure unexpectedly overcomes these shortcomings of the prior art.

[0088] Accordingly, this disclosure provides a method, system, and apparatus for detecting SNVs of target genes in samples with high accuracy, ensuring that only highly definitive negative samples are used to estimate noise, thereby obtaining more accurate background noise estimates.

[0089] Compared with the prior art, the present invention has at least the following advantages:

[0090] 1) Compared to PON composed of normal human samples, the clinical samples used in this invention are more representative of the complexity of tumor samples;

[0091] 2) Existing clustering methods cannot completely remove true variations, and may contain low-frequency variations. The threshold selection method established in this invention ensures that only highly specific negative samples are used to estimate noise;

[0092] 3) The truncated beta binomial model established in this invention is more flexible and more suitable for noise estimation across samples;

[0093] 4) According to the present invention, the testing provider can update the background noise assessment model in a timely manner based on the accumulation of clinical samples.

[0094] In the first aspect, this disclosure provides a method for detecting samples with high accuracy. i A device for detecting the SNV of a target gene in a genome, the device comprising: a sequencing data reading module for performing step (A) quality control on the raw sequencing data, aligning it to a reference genome and removing duplicates, and generating a BAM file that stores sequencing read information (sequence, position, quality, etc.);

[0095] The VAF calculation module is used to perform step (B) extracting the number of reads carrying variations at base sites on the genomic sequence from the deduplicated BAM file (x). i (i.e., reads whose bases at that position are inconsistent with the reference genome) and sequencing depth (n) i ), through x i / n i Calculate the variation frequency (VAF);

[0096] The threshold determination module is used to implement step (C) to calculate the threshold c using the Gaussian smoothing kernel function of the VAF distribution of the site, and classify samples that do not exceed c as negative;

[0097] The background noise estimation module is used to implement step (D) of selecting negative samples.i and n i The background noise parameter θ is estimated by fitting a truncated beta-binomial distribution and using the maximum likelihood method.

[0098] The SNV determination module is used to implement step (E) to determine the number of variant reads (x). i A threshold is set, and samples exceeding the threshold are identified as variants.

[0099] In one implementation, only when x i / n i When c ≤ c, there is no true variation in the samples used to estimate background noise.

[0100] In one implementation, in step (C), the threshold c is set as follows: (The remaining text appears to be incomplete and requires further context.) t i = log ( x i / n i ),right t i Apply Gaussian kernel smoothing to the distribution:

[0101]

[0102] in m For the sample size, s 2 The variance of the Gaussian kernel function. exp () is an exponential function. K ( t () is the kernel smoothed t i The distribution function;

[0103] right K ( t Find the maximum value and local minimum value respectively:

[0104]

[0105] Select VAF < t l For the samples, set a threshold c This is the 80th percentile of the VAF for this portion of the sample.

[0106] In one implementation, step (D) is performed as follows: for all negative samples (including VAF ≤ 0.5%) c and VAF > c )of x i and n i Establish a statistical model:

[0107]

[0108] in Let be the probability mass function of the beta-binomial. For the two-dimensional parameters of the model, For beta functions, ! represents the gamma function, and ! represents the factorial sign.

[0109]

[0110] Other notes The cumulative distribution function of the beta binomial:

[0111]

[0112] Exclude step (C) x i / n i > c After clinical samples, the number of variations The statistical distribution becomes a right-truncated beta binomial distribution:

[0113]

[0114] Estimating background noise parameters using the maximum likelihood method : Let be the log-likelihood function, when hour, Take the maximum value. This is the maximum likelihood estimate of the background noise, and Determine the distribution of background noise.

[0115] .

[0116] In one implementation, step (E) is performed as follows: for the tested sample ( x , n ),when p The value is less than the set false positive rate threshold. α The sample was determined to be positive at that time.

[0117] .

[0118] In one embodiment, the sample is a tumor sample, and the tumor is selected from solid tumors.

[0119] In one embodiment, the tumor is selected from pan-cancer solid tumors, including ovarian cancer, breast cancer, colorectal cancer, lung cancer, prostate cancer, liver cancer, stomach cancer, pancreatic cancer, cervical cancer, bladder cancer, kidney cancer, melanoma, and head and neck cancer.

[0120] In one implementation, the target gene is any gene contained in the original sequencing data.

[0121] In a more specific embodiment, the target gene is the BRAF gene, and the site is the V600E site.

[0122] Secondly, this disclosure provides a method for detecting samples with high accuracy. i Methods for SNV of target genes. (Reference) Figure 1 Step 1 corresponds to

[0123] Sequencing data reading step (A): Perform quality control on the raw sequencing data, align it to the reference genome and remove duplicates, and generate a BAM file that stores sequencing fragment (read) information;

[0124] VAF calculation step (B): Extract the number of reads carrying variations from the base sites on the genomic sequence (x) from the deduplicated BAM file. i ) and sequencing depth (n i ), through x i / n i Calculate the variation frequency (VAF);

[0125] Step 2 corresponds to

[0126] Threshold determination step (C): Calculate the threshold c using the Gaussian smoothing kernel function of the VAF distribution at this site, and classify samples that do not exceed c as negative;

[0127] Step 3 corresponds to

[0128] Background noise estimation step (D): Select x of the negative sample i and n i The background noise parameter θ is estimated by fitting a truncated beta-binomial distribution and using the maximum likelihood method.

[0129] Step 4 corresponds to

[0130] SNV determination step (E): Determine the number of mutated reads (x i A threshold is set, and samples exceeding the threshold are identified as variants.

[0131] In a more specific implementation, steps 1-4 are performed as follows:

[0132] 1) For suspected variant sites, from the sample iExtract the number of fragments containing the variant at this site from the deduplicated BAM file. x i and sequencing depth n i The variation frequency (VAF) is x i / n i .

[0133] 2) Because clinical samples contain a large number of real variant sites, including low-frequency variants, these need to be excluded before accurate calculation of background noise can be performed. Finding a threshold. c Only when x i / n i ≤ c Only then is it considered that the sample does not contain true variation; otherwise, there is a certain probability that variation exists. t i = log ( x i / n i ),right t i Apply Gaussian kernel smoothing to the distribution:

[0134]

[0135] in m For the sample size, s 2 The variance of the Gaussian kernel function. exp () is an exponential function.

[0136] right K ( t Find the maximum value and local minimum value respectively:

[0137]

[0138] Select VAF < t l For the samples, set a threshold c This is the 80th percentile of the VAF for this portion of the sample.

[0139] 3) To x i and n i Establish a statistical model:

[0140]

[0141] in Let be the probability mass function of the beta-binomial. For the two-dimensional parameters of the model, For beta functions, ! represents the gamma function, and ! represents the factorial sign.

[0142]

[0143] Other notes The cumulative distribution function of the beta binomial:

[0144]

[0145] Exclude 2) x i / n i > c After clinical samples, the number of variations The statistical distribution becomes a right-truncated beta-binomial distribution:

[0146]

[0147] Estimating background noise parameters using the maximum likelihood method . Let be the log-likelihood function, when hour, Take the maximum value. This is the maximum likelihood estimate of the background noise.

[0148]

[0149] 4) For the tested sample ( x , n ), when p The value is less than the set false positive rate threshold. α The sample was determined to be positive at that time.

[0150] .

[0151] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” used herein are also intended to include the plural forms. Furthermore, the open-ended expressions “comprising” and “including” are to be interpreted as potentially containing structural components or method steps not mentioned, but it should be noted that these open-ended expressions also cover situations where the invention consists only of the stated components and method steps (i.e., they cover the closed-ended expressions “consisting of…”).

[0152] As used throughout, a range is used as a shorthand to describe each and all values ​​within that range. Any value within a range, such as an integer value, a value incremented by one-tenth (when the range ends with one decimal place), or a value incremented by one-hundredth (when the range ends with two decimal places), can be chosen as the end of the range. For example, the range 0.1-10 is used to describe all values ​​within that range, such as 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8…9.5, 9.6, 9.7, 9.8, 9.9, and 10 (in increments of one-tenth), and includes all subranges, such as 0.1-1.0, 2.0-3.0, 4.0-5.0, 6.0-7.0, 8.0-9.0, etc.

[0153] All scientific and technical terms mentioned in this specification have the same meaning as commonly understood by those skilled in the art, and in case of conflict, the definitions in this specification shall prevail. To make the description of this invention easier to understand, some terms are explained below.

[0154] The term "high-throughput sequencing" used in this article, also known as next-generation sequencing (NGS) or massively parallel sequencing (MPS), refers to a sequencing technology that uses the principle of "sequencing while synthesis" to simultaneously perform parallel sequencing reactions on hundreds of thousands to millions of DNA molecules. The raw image data or electrochemical signals obtained are then analyzed through bioinformatics to ultimately obtain information such as the nucleic acid sequence or copy number of the sample. It is also called high-throughput sequencing, deep sequencing, or second-generation sequencing. The basic procedure of high-throughput sequencing involves randomly fragmenting the DNA to be tested into small fragments, constructing a library through steps such as end repair, ligation of adapter sequences, and PCR, and finally sequencing using sequencers such as Illumina or Ion Torrent.

[0155] The term "mutation / variation" as used herein refers to a genetic alteration in the genome of an organism or cell. For example, a targeted mutation / variation can be a change relative to the germline of an organism, such as a cancer cell-specific change. Mutations / variations can include single nucleotide variants (SNVs), copy number variants (CNVs), single nucleotide polymorphisms (SNPs), insertions, deletions, and rearrangements (e.g., fusions).

[0156] The term “SNV (single nucleotide variant)” as used in this article refers to a somatic single nucleotide mutation.

[0157] The terms “cancer” and “tumor” used herein are used interchangeably and can refer to cells or tissues exhibiting spontaneous, unregulated growth, whereby said cells or tissues exhibit abnormal growth characterized by increased cell proliferation (e.g., significant cell proliferation). The target cells used for detection, analysis, or treatment in this application may include, but are not limited to, precancerous cells (e.g., benign cells), malignant cells, pre-metastatic cells, metastatic cells, and non-metastatic cells. In fact, cancers of various tissues are known to those skilled in the art, including solid tumors such as sarcomas, glioblastomas, melanomas, lymphomas, myelomas, etc., and circulating cancers such as leukemia. Examples of cancers include, but are not limited to, ovarian cancer, breast cancer, colon cancer, lung cancer, prostate cancer, hepatocellular carcinoma, gastric cancer, pancreatic cancer, cervical cancer, ovarian cancer, liver cancer, bladder cancer, urinary tract cancer, thyroid cancer, kidney cancer, carcinoma, melanoma, head and neck cancer, and brain cancer.

[0158] While various embodiments of the invention have been described above, it should be understood that they are provided by way of example only and not as limitations. Many changes to the disclosed embodiments may be made in accordance with the disclosure herein without departing from the spirit or scope of the invention. Therefore, the breadth and scope of the invention should not be limited by any of the embodiments described above.

[0159] All references mentioned herein are incorporated herein by reference. All publications and patent documents cited in this application are incorporated herein by reference for all purposes, and are cited as if they were individually cited. Example

[0160] Unless otherwise stated, the materials used in the embodiments herein are all submitted tumor samples, and the various specific experimental methods used to conduct the experiments are all conventional experimental methods in the art or according to the steps and conditions recommended by the manufacturer, and can be routinely determined by those skilled in the art as needed.

[0161] Example 1: SNV detection of the V600E site of the BRAF gene in pan-cancer solid tumor samples

[0162] The protein encoded by the murine sarcoma-filterable bacteriogenic carcinogen homolog (BRAF) gene consists of 783 amino acids. The V600E mutation is a single thymine mutation in exon 15 that is replaced by adenine, resulting in the substitution of valine (V) for glutamate (E) at codon 600 of the protein. The BRAF V600E mutation has a significant impact on the occurrence, development, and prognosis of various tumors.

[0163] The method provided in this invention is used to detect BRAF V600E SNVs. The specific process is as follows:

[0164] (1) Run the sequencing data reading module to perform quality control on the raw sequencing data, align with the reference genome, remove duplicates, and generate a BAM file. Run the VAF calculation module to extract the number of variant reads (x) covering the BRAF V600E site from the BAM file. i ) and total number of reads (n i 6580 data points were obtained from 6580 pan-cancer solid tumor samples. x,n ): (0,460), (1,1105), (0,537), ....

[0165] (2). The threshold determination module is run to calculate the negative threshold c using a Gaussian smoothing kernel function of the x / n distribution. There are 392 negative samples out of 6580 samples. x > 0: (1,1105), (67,240), (220,661), ... This part of the sample ( x,n Substitute into the calculation .when hour, Take the maximum value, therefore .when hour Take the local minimum, therefore Choose non-zero values ​​that are less than 1. of x / n The quantiles are: 1 / 1105, 1 / 129, 1 / 223, ... The 80th percentile is... The samples with x / n ≤ c are the negative samples selected for modeling.

[0166] (3). Run the background noise estimation module and select the x-value of the negative sample. i and n i The background noise parameters are estimated by fitting a cutoff distribution. There are a total of 6499 samples. x / n ≤ c This part of the sample ( x,n )and Substitute ,when and hour Take the maximum value, therefore .

[0167] (4). Run the SNV determination module, using... Calculate the value of each sample Value, and set a threshold. α for For sample (1, 1105), The sample was determined to be negative. For sample (1,1105), The result was negative. For another sample (6,869), The sample was determined to be positive, and the positive sample was verified by a third-party experimental platform to be a genuine mutant sample.

[0168] The data from this embodiment demonstrate that, compared to PON composed of normal human samples, the present invention achieves more accurate noise estimation and accurate interpretation of low-frequency variations using clinical samples. Furthermore, the data from this embodiment also show that the method is applicable to a wide range of cancerous solid tumor samples, and is not limited to any specific tumor type; that is, various solid tumor types can benefit from the method and apparatus of the present invention.

[0169] While various embodiments of the invention have been described above, it should be understood that they are provided by way of example only and not as limitations. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications will fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting samples with high accuracy i A device for detecting the SNV of a target gene in the middle, the device comprising: The sequencing data reading module is used to perform quality control on the raw sequencing data in step (A), align it to the reference genome and remove duplicates, and generate a BAM file that stores sequencing fragment (read) information; The VAF calculation module is used to perform step (B) extracting the number of reads carrying variations at base sites on the genomic sequence from the deduplicated BAM file (x). i ) and sequencing depth (n i ), through x i / n i Calculate the Variant Allele Frequency (VAF); The threshold determination module is used to implement step (C) to calculate the threshold c using the Gaussian smoothing kernel function of the VAF distribution of the site, and classify samples that do not exceed c as negative; The background noise estimation module is used to implement step (D) of selecting negative samples. i and n i The background noise parameter θ is estimated by fitting a truncated beta-binomial distribution and using the maximum likelihood method. The SNV determination module is used to implement step (E) to determine the number of variant reads (x). i A threshold is set, and samples exceeding the threshold are classified as variants. Where only if x i / n i When c ≤ c, there is no true variation in the samples used to estimate background noise. In step (C), the threshold c is set as follows: [The text is incomplete and requires further context to be accurately translated.] t i = log ( x i / n i ),right t i Apply Gaussian kernel smoothing to the distribution: in m For the sample size, σ 2 Let the variance be the Gaussian kernel function. exp () is an exponential function. K ( t () is the kernel smoothed t i The distribution function; right K ( t Find the maximum value and local minimum value respectively: Select VAF < t l For the samples, set a threshold c This is the 80th percentile of the VAF for this portion of the sample. Step (D) is performed as follows: for VAF ≤ c and VAF > c All negative samples x i and n i Establish a statistical model: in Let be the probability mass function of the beta binomial. For the two-dimensional parameters of the model, For beta functions, ! represents the gamma function, and ! represents the factorial sign. Other notes The cumulative distribution function of the beta binomial: Exclude step (C) x i / n i > c After clinical samples, the number of variations The statistical distribution becomes a right-truncated beta binomial distribution: Estimating background noise parameters using the maximum likelihood method : Let be the log-likelihood function, when hour, Take the maximum value. This is the maximum likelihood estimate of the background noise, and Determine the distribution of background noise. ， Step (E) is performed as follows: For the tested sample ( x , n ),when p The value is less than the set false positive rate threshold. α The sample was determined to be positive at that time. ， The sample mentioned is a tumor sample.

2. The device according to claim 1, wherein the tumor is selected from solid tumors.

3. The device according to claim 2, wherein the tumor is a pan-cancer solid tumor.

4. The device according to claim 3, wherein the pan-cancer solid tumor is selected from ovarian cancer, breast cancer, colorectal cancer, lung cancer, prostate cancer, liver cancer, stomach cancer, pancreatic cancer, cervical cancer, bladder cancer, kidney cancer, melanoma, and head and neck cancer.

5. The apparatus according to any one of claims 1-4, wherein the target gene is any gene contained in the original sequencing data.

6. The apparatus according to any one of claims 1-4, wherein the target gene is the BRAF gene and the site is the V600E site.

Images