Gene mutation detection method, device, equipment, medium and product

Abstract

The invention discloses a gene mutation detection method, device and equipment, a medium and a product. The method comprises the steps that suspected mutation sites of a nucleic acid sample to be detected are obtained, and the suspected mutation sites are determined through first mutation characteristic data obtained by conducting mutation detection on sequencing data of the nucleic acid sample to be detected through a first mutation detection module; the recall rate of a gene mutation site identified by the first mutation detection module is greater than or equal to a preset recall rate; acquiring second mutation feature data and third mutation feature data of the suspected mutation sites, inputting the second mutation feature data and the third mutation feature data into a pre-trained target mutation detection model, and outputting mutation detection results of the suspected mutation sites.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biotechnology, and in particular to a gene mutation detection method, device, equipment, medium and product. BACKGROUND

[0002] Gene mutation detection is one of the important links in the process of bioinformatics data processing. Due to the complexity of biological data, affected by factors such as sequencing bias of sequencing platform and batch bias of sequencing reagent, the current gene mutation detection method or software system still has certain detection error.

[0003] Especially for mutation sites with low mutation frequency, such as mutation frequency less than 1.5%, the current gene mutation detection method or software system cannot guarantee the precision and recall rate of gene mutation detection at the same time. SUMMARY

[0004] The embodiments of the present application provide a gene mutation detection method, device, equipment, medium and product, which are used to solve the problem of poor performance of traditional mutation detection software, so as to balance the precision and recall rate of gene mutation detection.

[0005] According to an embodiment of the present application, a gene mutation detection method is provided, which comprises:

[0006] Obtaining suspected mutation sites of a nucleic acid sample to be detected, wherein the suspected mutation sites are determined by using first mutation detection module to obtain first mutation feature data of the sequencing data of the nucleic acid sample to be detected, and the recall rate of the first mutation detection module to identify gene mutation sites is greater than or equal to a preset recall rate;

[0007] Using a second mutation detection module, extracting features of each of the suspected mutation sites to obtain second mutation feature data;

[0008] Using a sequencing data processing module to process the sequencing data to obtain total sequencing processing data, and screening sequencing processing data of each of the suspected mutation sites in the total sequencing processing data to obtain third mutation feature data;

[0009] Inputting the second mutation feature data and the third mutation feature data into a pre-trained target mutation detection model to output mutation detection results of each of the suspected mutation sites.

[0010] According to another embodiment of the present application, a gene mutation detection device is provided, which comprises:

[0011] The suspected mutation site acquisition module is configured to acquire suspected mutation sites of a nucleic acid sample to be detected, wherein the suspected mutation sites are determined by first mutation feature data obtained by using a first mutation detection module to detect mutations of sequencing data of the nucleic acid sample to be detected, and the first mutation detection module has a recall rate of identifying gene mutation sites greater than or equal to a preset recall rate;

[0012] The second mutation feature data acquisition module is configured to acquire second mutation feature data by using a second mutation detection module to extract features of each suspected mutation site;

[0013] The third mutation feature data acquisition module is configured to acquire third mutation feature data by using a sequencing data processing module to process the sequencing data to obtain total sequencing processing data, and to filter sequencing processing data of each suspected mutation site in the total sequencing processing data;

[0014] The mutation detection result output module is configured to input the second mutation feature data and the third mutation feature data into a target mutation detection model that has been pre-trained, and to output mutation detection results of each suspected mutation site.

[0015] According to another embodiment of the present application, an electronic device is provided, which comprises:

[0016] at least one processor; and

[0017] a memory connected to the at least one processor in communication; wherein

[0018] the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the method for detecting gene mutations according to any one of the embodiments of the present application.

[0019] According to another embodiment of the present application, a computer readable storage medium is provided, which stores computer instructions for enabling a processor to implement the method for detecting gene mutations according to any one of the embodiments of the present application when executed by the processor.

[0020] According to another embodiment of the present application, a computer program product is provided, which comprises a computer program for implementing the method for detecting gene mutations according to any one of the embodiments of the present application when executed by a processor.

[0021] The technical scheme of the embodiment of the present application acquires the suspected mutation site of the nucleic acid sample to be detected by using the first mutation detection module with the recall rate greater than or equal to the preset recall rate, thereby ensuring the integrity of the suspected mutation site detected; on this basis, the second mutation detection module is used to extract the second mutation feature data from the suspected mutation site, the sequencing data processing module is used to process the sequencing data of the suspected mutation site, the total sequencing processing data is obtained, the sequencing processing data of each suspected mutation site in the total sequencing processing data is screened, and the third mutation feature data is obtained; the second mutation feature data and the third mutation feature data are input into the target mutation detection model pre-trained, and the mutation detection result of each suspected mutation site is output. The target mutation detection model provided by the present application is based on the first mutation feature data with high recall rate contribution, and combines the second mutation feature data extracted by the second mutation detection module different from the first mutation feature detection module and the third mutation feature data associated with the suspected mutation feature screened in the sequencing data, so that the combination of the triple mutation features solves the problem of poor detection performance of the traditional mutation detection method or software system, and the high accuracy and high recall rate of the gene mutation detection are taken into account.

[0022] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0024] Figure 1 A flowchart of a gene mutation detection method provided by an embodiment of the present application;

[0025] Figure 2 A flowchart of another gene mutation detection method provided by an embodiment of the present application;

[0026] Figure 3 A meaning diagram of an average alignment quality value provided by an embodiment of the present application;

[0027] Figure 4 A diagram of unique germline mutation sites of HG002 relative to HG001 provided by an embodiment of the present application;

[0028] Figure 5A schematic diagram of a unique germline mutation site of HG002 relative to a certain site of HG001 provided by an embodiment of the present application;

[0029] Figure 6 A schematic diagram of a method for constructing sequencing data simulating somatic mutations provided by an embodiment of the present application;

[0030] Figure 7 A schematic diagram of a training method of an xgboost model provided by an embodiment of the present application;

[0031] Figure 8 A structural schematic diagram of a gene mutation detection device provided by an embodiment of the present application;

[0032] Figure 9 A schematic diagram of a tumor standard sample sequencing data construction module provided by an embodiment of the present application;

[0033] Figure 10 A structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative labor should fall within the scope of protection of the present application.

[0035] It should be noted that the terms "first", "second", "third", "initial", "target", "reference" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0036] In embodiments of the present application, the term "sequencing" can also be referred to as "nucleic acid sequencing" or "genetic sequencing", all of which are interchangeable in expression and refer to the determination of the type and order of bases or nucleotides (including nucleotide analogs) in a nucleic acid molecule. The term sequencing includes the process of binding nucleotides to a template and collecting the corresponding signals emitted by the nucleotides (including analogs). The term sequencing includes sequencing by synthesis (SBS) and / or sequencing by ligation (SBL), including DNA sequencing and / or RNA sequencing; the sequencing results / data obtained by sequencing, i.e., the nucleic acid fragments or read sequences read out, are referred to as reads.

[0037] In embodiments of the present application, VarScan is a software, specifically a java-written software for tumor somatic mutation detection (calls SNV in somatic variants) under a linux system, which can be used for targeted, exome and whole-genome resequencing and other sequencing data. Strelka2 is a mutation detection software for analyzing somatic variants of small cohorts and tumor / no mutation sample pairs. Mutect2 is a tool in GATK (The Genome Analysis Toolkit, a software developed by Broad Institute for analysis of second-generation resequencing data), which mainly functions to detect somatic variants, including SNV and InDel.

[0038] When detecting gene mutations of biological samples, the performance of traditional mutation detection software is limited, and there is a problem that it is difficult to balance the precision and recall of mutation detection. In particular, for biological samples with low mutation frequency, the defects of model performance are particularly obvious.

[0039] In the present application, the term "recall" is also referred to as "recall rate", which is used to represent how many actual positive samples are correctly predicted. Specifically, the recall rate is represented as the proportion of the number of samples predicted as positive samples and actually as positive samples to the total number of actual positive samples, or the probability of positive samples predicted as positive samples by the model. For example, the recall rate can be represented as Recall = TP / (TP+FN).

[0040] The term "precision" is also referred to as "positive predictive value" and is used to represent how many of the samples predicted as positive samples are actually positive samples. The precision is expressed as the number of samples predicted as positive samples and actually positive samples divided by the total number of samples predicted as positive samples, or the probability of the samples predicted as positive samples actually being positive samples. For example, the precision can be expressed as Precision = TP / (TP+FP).

[0041] wherein TP represents the number of instances correctly predicted as positive class by the model, or the number of samples predicted as positive samples and actually positive samples by the model. FP represents the number of instances of actual negative class incorrectly predicted as positive class by the model, or the number of samples predicted as positive samples but actually negative samples. FN represents the number of instances of actual positive class incorrectly predicted as negative class by the model, or the number of samples predicted as negative samples but actually positive samples. Taking the detection of genetic mutations as an example, TP represents the number of genetic sites predicted by the model as having genetic mutations and actually having genetic mutations. FP represents the number of genetic sites predicted by the model as having genetic mutations but actually not having genetic mutations. FN represents the number of genetic sites predicted by the model as not having genetic mutations but actually having genetic mutations. In the embodiments of the present application, the positive samples are real mutation sites, and the negative samples are false positive mutation sites. The false positive mutation sites can represent variation sites caused by sequencing errors.

[0042] The following describes biological samples with a mutation frequency of 0.5% to 1.5% as research objects.

[0043] When Strelka 2 and Mutect 2 are used for genetic mutation detection, as the sequencing depth increases, the recall rate will reach a plateau value. At this time, even if the sequencing depth is increased, there is no significant increase in the recall rate. This plateau value is related to the sequencing platform, especially the error rate of the sequencing. For example, the recall rate of the Element AVITI platform is about 0.85, and the recall rate of the NovaSeq 6000 platform is about 0.75. The precision of genetic mutation detection is determined by the sequencing platform and can basically maintain a high precision of greater than 0.9.

[0044] When VarScan is used for genetic mutation detection, compared with Strelka 2 and Mutect 2, as the sequencing depth increases, a high recall rate of about 0.96 can be reached. However, VarScan has the disadvantage of extremely low precision. In actual application, in order to improve the accuracy, it is necessary to rely on genome annotation, genetic annotation, manual interpretation and other methods to identify real mutations after mutation detection, which has low automation degree and is extremely time-consuming and laborious.

[0045] Therefore, when facing somatic mutations with mutation frequency <1.5%, neither the sequencing platform nor the increase in sequencing depth can achieve high recall rate and high precision rate at the same time.

[0046] Therefore, the embodiments of the present application provide a gene mutation detection method, device, equipment, medium and product, which can obtain better gene mutation detection results, and can balance high precision rate and high recall rate even for gene mutations with mutation frequency less than or equal to 1.5%.

[0047] Figure 1 A flowchart of a gene mutation detection method provided by an embodiment of the present application, the embodiment can be applicable to the case of mutation detection on the variant sites in the sequencing data, and the method can be executed by a gene mutation detection device which can be realized in the form of hardware and / or software and can be configured in a terminal device.

[0048] As shown in Figure 1 , the method comprises:

[0049] S110, acquiring a suspected mutation site of a nucleic acid sample to be detected.

[0050] Specifically, the nucleic acid sample to be detected is used to represent a sample containing nucleic acid molecules or a nucleic acid mixture and used for mutation detection, wherein the nucleic acid mixture includes two or more nucleic acid molecules with different sequences. In the embodiments of the present application, the term "nucleic acid molecule" represents a polymer form of nucleotides with any length, which can be a nucleic acid short fragment with a length less than or equal to 500 bp, for example, a nucleic acid fragment with a length of 20 bp, 50 bp, 80 bp, 100 bp, 120 bp, 150 bp, 180 bp, 200 bp, 220 bp, 250 bp, 280 bp, 300 bp, 320 bp, 350 bp, 380 bp, 400 bp, 420 bp, 450 bp, 480 bp, 500 bp, etc.; or a nucleic acid long fragment with a length greater than 500 bp. In the embodiments of the present application, the "nucleic acid molecule" includes a nucleic acid molecule composed of ribonucleotides or a nucleic acid molecule composed of deoxyribonucleotides. In addition, the nucleic acid sample to be detected in the embodiments of the present application can be a nucleic acid sample after amplification, that is, for each sequence of the nucleic acid molecules in the nucleic acid sample to be detected, there are multiple copies of the nucleic acid molecules.

[0051] Exemplarily, the sample source includes, but is not limited to, blood, urine, plasma, a cell sample, an ex vivo tissue fluid or a tumor tissue, and the like, and the sample source of the nucleic acid sample to be detected is not limited herein. In an embodiment, the sample is a nucleic acid sample to be detected from a human, that is, the method for detecting a gene mutation provided in the embodiments is a method for detecting a gene mutation of a human sample. It should be understood that the method is also applicable to detecting a gene mutation of a nucleic acid sample to be detected from other biological species. In the following description, the nucleic acid sample to be detected from a human is taken as an example for illustration, but is not used to limit the species source of the nucleic acid sample to be detected.

[0052] In some embodiments, the nucleic acid sample to be detected is a nucleic acid sample with a mutation frequency less than or equal to 1.5%. Specifically, the mutation frequency refers to, for a sample, the ratio of the sequencing depth of a specific mutation at a specific genomic site to the total sequencing depth. For example, in a reference genome, the total number of bases at genomic site P is 100 (i.e., the sequencing depth is 100) and the base is adenine A (i.e., if no mutation occurs, theoretically, the base detected at genomic site P is all adenine A). In the gene sequencing data of the nucleic acid sample to be detected, a cytosine C mutation occurs at genomic site P and the number of cytosine C bases is 5 (i.e., the sequencing depth is 5), and the gene mutation frequency of the nucleic acid sample to be detected is 5%.

[0053] The conventional gene mutation detection software has poor detection performance for a nucleic acid sample with a mutation frequency less than or equal to 1.5%, and cannot simultaneously ensure high accuracy and high recall rate. The target mutation detection model in the embodiments is particularly suitable for detecting a nucleic acid sample with a mutation frequency less than or equal to 1.5%, and can simultaneously consider high accuracy and high recall rate of the detection result.

[0054] In some embodiments, the nucleic acid sample to be detected is a biological sample with a somatic mutation, or the nucleic acid sample to be detected is a biological sample with a germ cell mutation. The somatic mutation refers to a mutation occurring in a somatic cell during individual development due to the influence of mutagenic factors, and the mutation is usually not inherited to the next generation; the germ cell mutation refers to a mutation carried by a germ cell during sexual reproduction, and the mutation can be inherited to the next generation.

[0055] Specifically, the sequencing data of the nucleic acid sample to be detected is used to represent the nucleic acid sequence data obtained by sequencing the nucleic acid sample to be detected by using a sequencing technology. For example, the sequencing technology includes but is not limited to Sanger sequencing technology, high-throughput sequencing technology, and single molecule sequencing technology, and the like. Here, the sequencing technology is not limited, and can be defined according to actual needs. The sequencing mode can be single-end sequencing or double-end sequencing. Sequencing can be performed by a sequencing platform. The sequencing platform can be selected from, but is not limited to, Hiseq / Miseq / Nextseq / Novaseq sequencing platforms of Illumina company, Ion Torrent platform of ThermoFisher / Life Technologies company, BGISEQ and MGISEQ platforms of Huada Gene, and single molecule sequencing platform, and GenoLab M / GenoCare 1600 / FASTASeq 300 / SURFSec5000 sequencing platforms of Zmappia, and the like.

[0056] In the embodiments of the present application, the mutation site refers to a gene site in the nucleic acid sample, which is inconsistent with the reference genome due to base mutation. It should be understood that the mutation site in the embodiments of the present application belongs to the category of variation sites, and the ratio of the mutation site to the variation site should be between 0 and 1.

[0057] In the embodiments of the present application, the suspected mutation site is determined by the first mutation detection module for detecting the mutation of the sequencing data of the nucleic acid sample to be detected. By using the first mutation detection module to analyze the sequencing data of the nucleic acid sample to be detected, the mutation features contained in the sequencing data itself can be extracted, and / or the mutation features obtained by further data processing of the sequencing data. The mutation feature data obtained by the first mutation detection module is named as the first mutation feature data in the embodiments of the present application, that is, the suspected mutation site in the embodiments of the present application is determined by the first mutation feature data obtained by using the first mutation detection module to detect the mutation of the sequencing data of the nucleic acid sample to be detected.

[0058] In the embodiments of the present application, when the first mutation detection module for detecting the mutation of the nucleic acid sample to be detected identifies the gene mutation site, i.e., the mutation detection, the recall rate is greater than or equal to the preset recall rate, i.e., the recall rate of the first mutation detection module for identifying the gene mutation site is greater than or equal to the preset recall rate, so that the probability of identifying the real mutation site meets the preset probability requirement. Therefore, the suspected mutation site obtained by the first mutation detection module is subjected to the following step S120, so that the second mutation detection module with the preset accuracy requirement is further analyzed, and the second mutation feature data extracted by the second mutation detection module is used as one of the input features of the target mutation detection model, so that the target mutation detection model in step S140 can output the mutation detection result that can take into account the high recall rate and the high accuracy.

[0059] In some embodiments, the preset recall rate is 0.9, which means that the recall rate of the first mutation detection module for identifying the gene mutation site, i.e., the mutation detection, is greater than or equal to 0.9, thereby reducing the probability of missing the identification of the real mutation site. In some embodiments, the recall rate of the first mutation detection module for identifying the gene mutation site is greater than or equal to 0.95.

[0060] In some embodiments, the suspected mutation site of the nucleic acid sample to be detected is obtained by: aligning the sequencing data of the input nucleic acid sample to be detected with the sequencing data of the reference genome by the alignment unit in the first mutation detection module to obtain at least one gene variation site in the sequencing data of the nucleic acid sample to be detected that is different from the sequencing data of the reference genome; extracting features of the at least one gene variation site by the feature extraction unit in the first mutation detection module to obtain at least one first mutation feature data; and screening each gene variation site according to the at least one first mutation feature data by the mutation detection unit in the first mutation detection module to obtain the suspected mutation site.

[0061] The advantage of such setting is that, since the first mutation detection module has a high recall rate, it not only ensures the integrity of the identified suspected mutation site, but also filters out some false positive mutation sites, thereby reducing the data amount of the suspected mutation site and improving the detection efficiency of the gene mutation.

[0062] The variant site can be obtained by using a sequencing data alignment tool. For example, the alignment tool can be Bowtie2 or BLAST, and the alignment tool is not limited herein and can be customized according to actual needs.

[0063] The variant site can be obtained by using a sequencing data alignment tool. For example, the alignment tool can be Bowtie2 or BLAST, and the alignment tool is not limited herein and can be customized according to actual needs.

[0064] The mutation site refers to a gene site in the nucleic acid sample, in which a sequence is inconsistent with a reference genome due to a base mutation, including a single nucleotide variation and a single nucleotide variation. It should be understood that the mutation site in the embodiments of the present application belongs to the category of the variant site, and the ratio of the mutation site to the variant site should be between 0 and 1.

[0065] In S120, a second mutation detection module is used to extract second mutation feature data from the suspected mutation sites.

[0066] Here, in order to distinguish the first mutation feature data extracted by the first mutation detection module, the mutation feature data extracted by the second mutation detection module is named as “second mutation feature data”.

[0067] In the embodiments of the present application, the first mutation detection module and the second mutation detection module are both models for detecting mutations in nucleic acid samples, and the mutation detection principle of the second mutation detection module is different from that of the first mutation detection module. In the embodiments of the present application, the second mutation detection module is introduced, and the mutation detection function in the second mutation detection module is used to extract features from the suspected mutation sites from the first mutation detection module, so as to improve the accuracy of the target mutation detection model trained by taking the extracted features as one of the characteristics, and to make the proportion of the real mutation sites in the suspected mutation sites recognized by the first mutation detection module meet the preset proportion requirement.

[0068] In some embodiments, the precision of the second mutation detection module is greater than or equal to a preset precision, and the preset precision satisfies: the precision of the target mutation detection model is greater than or equal to 0.8. At this time, by setting the first mutation detection module that satisfies the preset recall rate requirement, the probability that the true mutation site is identified can be improved, and then the suspected mutation site obtained by the first mutation detection module is further analyzed by the second mutation detection module that satisfies the preset precision requirement, and the extracted second mutation feature is used as one of the input features of the target mutation detection model, so that the mutation detection result output by the target mutation detection model in step S140 can be a mutation detection result with high recall rate and high precision.

[0069] In some embodiments, the preset precision is 0.9, which means that the precision of the second mutation detection module in identifying the gene mutation site, i.e., the mutation detection, is greater than or equal to 0.9. At this time, in identifying the gene mutation site, the first mutation detection module has a recall rate greater than or equal to the preset recall rate, and the second mutation detection module has a precision greater than or equal to the preset precision. Therefore, by setting the first mutation detection module that satisfies the preset recall rate requirement, the probability that the true mutation site is identified can be improved, and then the suspected mutation site obtained by the first mutation detection module is further analyzed by the second mutation detection module that satisfies the preset precision requirement, and the extracted second mutation feature is used as one of the input features of the target mutation detection model, so that the mutation detection result output by the target mutation detection model in step S140 can be a mutation detection result with high recall rate and high precision.

[0070] In some embodiments, the recall rate of the first mutation detection module in identifying the gene mutation site, i.e., the mutation detection, is greater than or equal to 0.9; and the precision of the second mutation detection module in identifying the gene mutation site, i.e., the mutation detection, is greater than or equal to 0.9. Therefore, by first reducing the probability of missing identification of the true mutation site through the first mutation detection module, the probability of identifying the suspected mutation site is improved, and then the suspected mutation site obtained by the first mutation detection module is further analyzed by the second mutation detection module to reduce the probability of missing identification of the false positive mutation site, and finally the second mutation feature data extracted by the second mutation detection module is used as one of the input features of the target mutation detection model, so that the mutation detection result output by the target mutation detection model in step S140 can have high recall rate and high precision.

[0071] In the embodiments of the present application, the recall rate and the precision can be obtained by testing the first mutation detection module and the second mutation detection module through a test set. The test set contains test sequencing data of test nucleic acid samples.

[0072] In an embodiment, the first mutation detection module and the second mutation detection module can be obtained by screening from the currently used gene mutation detection software. For example, the module architecture of the first mutation detection module is Varscan software, and the module architecture of the second mutation detection module is Mutect2 software. At this time, when identifying the gene mutation site, the recall rate of the Varscan software is high, which can reach 0.96, thereby improving the probability of identifying the true mutation site. The suspected mutation site obtained by the Varscan software is further analyzed by the Mutect2 software with a precision greater than or equal to 0.9, and the second mutation feature data obtained is used as one of the input features of the target mutation detection model together with the third mutation feature data extracted from the sequencing data for the suspected mutation site. When the third mutation feature data is extracted from the sequencing data for the suspected mutation site, the mutation detection result output by the target mutation detection model in step S140 can take into account the mutation detection result with high recall rate and high precision. The results show that the detection recall rate and precision of the target mutation detection model for the mutation site are both optimal.

[0073] S130, processing the sequencing data by using the sequencing data processing module to obtain total sequencing processing data, and screening the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain third mutation feature data.

[0074] In some embodiments, screening the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain the third mutation feature data comprises: for each suspected mutation site, obtaining the sequencing processing data corresponding to the suspected mutation site in the total sequencing processing data, and adding the sequencing processing data to the third mutation feature data.

[0075] In some embodiments, screening the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain the third mutation feature data comprises: for each suspected mutation site, obtaining the sequencing processing data corresponding to the suspected mutation site in the total sequencing processing data; performing feature screening on the sequencing processing data, and adding the screened sequencing processing data to the third mutation feature data.

[0076] S140, inputting the second mutation feature data and the third mutation feature data into the target mutation detection model pre-trained to output the mutation detection result of each suspected mutation site.

[0077] In the embodiments of the present application, the second mutation feature data and the third mutation feature data are as described above.

[0078] In the embodiments of the present application, the representation form of the mutation detection result can be whether it is a positive site, a level, a score, or a probability, etc. The level can be a true mutation level or a false positive mutation level, the score can be a true mutation score or a false positive mutation score, and the probability can be a true mutation rate or a false positive mutation rate, etc. Taking the true mutation level, the true mutation score, or the true mutation rate as an example, the higher the true mutation level, the true mutation score, or the true mutation rate, the greater the possibility that the suspected mutation site is a true mutation site, and vice versa, the lower the true mutation level, the true mutation score, or the true mutation rate, the smaller the possibility that the suspected mutation site is a true mutation site.

[0079] Here, the representation form of the target mutation detection result is not limited, and can be set according to actual needs.

[0080] In an optional embodiment, the mutation detection result is a result of the suspected mutation site being a positive site.

[0081] In an optional embodiment, the mutation detection result is a result of the suspected mutation site being a negative site.

[0082] The result of the positive site can be confirmed by the true mutation level, the true mutation score, and the true mutation rate, but the index for confirming that the suspected mutation site is a positive site is not limited thereto. The result of the negative site can be confirmed by the false positive mutation level, the false positive mutation score, and the false positive mutation rate, and similarly, the index for confirming that the suspected mutation site is a negative site is not limited thereto.

[0083] In some embodiments, the second mutation feature data and the third mutation feature data are input into a pre-trained target mutation detection model, and a mutation detection result of each suspected mutation site is output, including:

[0084] The second mutation feature data and the third mutation feature data are input into a pre-trained target mutation detection model, and a predicted value of the suspected mutation site being positive is obtained;

[0085] The predicted value is compared with a preset target value, and a comparison result is output:

[0086] If the predicted value is greater than the preset target value, a result of the suspected mutation site being a positive site is output;

[0087] If the predicted value is less than the preset target value, a result of the suspected mutation site being a negative site is output.

[0088] The technical scheme of the embodiment, by adopting the first mutation detection module with a recall rate greater than or equal to a preset recall rate to obtain the suspected mutation site of the nucleic acid sample to be detected, ensures the integrity of the suspected mutation site detected; on this basis, the second mutation detection module is used to extract the second mutation feature data from the suspected mutation site, and the sequencing data processing module is used to process the sequencing data of the suspected mutation site, obtain the total sequencing processing data, and screen the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain the third mutation feature data; the second mutation feature data and the third mutation feature data are input into the target mutation detection model pre-trained to output the mutation detection result of each suspected mutation site. The target mutation detection model provided by the application combines the second mutation feature data extracted by the second mutation detection module different from the first mutation feature detection module and the third mutation feature data associated with the suspected mutation feature screened from the sequencing data on the basis of the first mutation feature data with a high recall rate, and the combination of the three mutation features solves the problem of poor detection performance of the traditional mutation detection method or software system, and takes into account the high accuracy and high recall rate of gene mutation detection.

[0089] Figure 2 The flowchart of another gene mutation detection method provided by an embodiment of the application, the module architecture of the first mutation detection module in the embodiment is Varscan software, the module architecture of the second mutation detection module is Mutect2 software or Strelka2 software, and the sequencing data processing module is bam-readcount software. As shown in the figure, the method comprises: Figure 2

[0090] S210, obtaining the suspected mutation site of the nucleic acid sample to be detected determined by the Varscan software.

[0091] In the embodiment, the nucleic acid sample to be detected is a nucleic acid sample to be detected from somatic cells, and the suspected mutation site is determined by the first mutation feature data obtained by mutation detection on the sequencing data of the nucleic acid sample to be detected by the Varscan software.

[0092] The VarScan software is a software for somatic mutation detection under a java linux system. With the increase of sequencing depth, the recall rate of the VarScan software can reach about 0.96, but the accuracy of the VarScan software is poor. In the prior art, after mutation detection by the VarScan software, further verification of the mutation detection result output by the VarScan software is still needed by relying on methods such as genome annotation, genetic annotation and manual judgment.

[0093] ​S220, acquiring second mutation feature data of the suspected mutation site extracted by the Mutect2 software or the Strelka2 software, and third mutation feature data of the suspected mutation site extracted by the bam-readcount software.

[0094] In step S220, the second mutation feature data is feature data obtained by extracting features of the suspected mutation site determined by the VarScan software using the Mutect2 software or the Strelka2 software. The Mutect2 software or the Strelka2 software is software for detecting somatic mutations. The Mutect2 software is based on a Bayesian algorithm of a hidden Markov model and mainly detects single nucleotide site variations and insertion-deletion variations of somatic cells. The Strelka2 software rearranges gene variation sites of insertion-deletion variations, and detects mutations based on a Bayesian probability model and mutation feature data of the rearranged gene variation sites. With the increase of sequencing depth, the recall rate of the Mutect2 software or the Strelka2 software reaches a platform value. The platform value is affected by the average error rate of the sequencing platform to which the sequencing technology belongs, so the platform value is not very high, but the precision rate of the Mutect2 software or the Strelka2 software can reach more than 0.9.

[0095] In some embodiments, the second mutation feature data includes at least one of sequencing depth, number of variation events, germline variant quality score, median reference base quality value, median mutant base quality value, median reference base insertion fragment length, median mutant base insertion fragment length, median reference base alignment quality value, median mutant base alignment quality value, distance median, no mutation sample false positive log odds, no mutation sample log likelihood ratio, and disease log likelihood ratio.

[0096] In some embodiments, the second mutation feature data includes at least one of sequencing depth, number of variation events, germline variant quality score, median reference base quality value, median mutant base quality value, median reference base insertion fragment length, median mutant base insertion fragment length, median reference base alignment quality value, median mutant base alignment quality value, distance median, no mutation sample false positive log odds, no mutation sample log likelihood ratio, and disease log likelihood ratio.

[0097] In some embodiments, the second mutation feature data includes at least one of sequencing depth, number of variation events, germline variant quality score, median reference base quality value, median mutant base quality value, median reference base insertion fragment length, median mutant base insertion fragment length, median reference base alignment quality value, median mutant base alignment quality value, distance median, no mutation sample false positive log odds, no mutation sample log likelihood ratio, and disease log likelihood ratio.

[0098] wherein the Read Depth (DP) is used to represent the number of times a locus is covered by reads. For example, DP = 4656 means that 4656 reads cover the locus in the obtained sequencing data.

[0099] wherein the Number of Events (ECNT) represents the number of observed variant events for the identified suspected mutation locus. The variant events include various situations that make the base at the locus of the reads different from the standard base of the human reference genome, including but not limited to variant events caused by gene insertion, gene deletion, gene mutation, etc.

[0100] wherein the Germline Quality (GERMQ) represents the quality score of the identified suspected mutation locus not belonging to the germline variation, and is used to represent the possibility of the suspected mutation locus not belonging to the germline variation. The higher the germline quality score, the higher the possibility of the suspected mutation locus not belonging to the germline variation. For example, assuming that GERMQ = 93, it means that the suspected mutation locus has a high possibility of not belonging to the germline variation.

[0101] wherein the Median Base Quality (MBQ) represents the median of the quality values of the base corresponding to the suspected mutation locus in the reference genome. In some examples, the median base quality includes the median of the base quality values of the mutant allele (ALT) corresponding to the suspected mutation locus in the reads. In some examples, the median base quality includes the median of the base quality values of the reference allele (REF) corresponding to the suspected mutation locus in the reference genome. In some examples, the median base quality includes the median of the base quality values of the mutant allele (ALT) corresponding to the suspected mutation locus in the reads, and the median of the base quality values of the reference allele (REF) corresponding to the suspected mutation locus in the reference genome. For example, MBQ = 20, 20 means that the median of the base quality values of the mutant allele (ALT) and the reference allele (REF) of the locus are both 20.

[0102] wherein the Median Base Quality (MBQ) represents the median of the quality values of the base corresponding to the suspected mutation locus in the reference genome. In some examples, the median base quality includes the median of the base quality values of the mutant allele (ALT) corresponding to the suspected mutation locus in the reads. In some examples, the median base quality includes the median of the base quality values of the reference allele (REF) corresponding to the suspected mutation locus in the reference genome. In some examples, the median base quality includes the median of the base quality values of the mutant allele (ALT) corresponding to the suspected mutation locus in the reads, and the median of the base quality values of the reference allele (REF) corresponding to the suspected mutation locus in the reference genome. For example, MBQ = 20, 20 means that the median of the base quality values of the mutant allele (ALT) and the reference allele (REF) of the locus are both 20.

[0103] wherein the median fragment length by allele (MFRL) indicates the median of the insert fragment length of the paired read sequence at the same base as the reference genome base for the suspected mutation site. In some examples, the median fragment length includes the median fragment length of the reference allele (REF) in the reference genome for the suspected mutation site. In some examples, the median fragment length includes the median fragment length of the mutant allele (ALT) in the read sequence for the suspected mutation site, and the median fragment length of the reference allele (REF) in the reference genome for the suspected mutation site. For example, MFRL = 199, 191 indicates that the median fragment length of the mutant allele and the reference allele are 199 and 191, respectively.

[0104] wherein the median fragment length by mutant (MFLM) indicates the median of the insert fragment length of the paired read sequence at the mutant base for the suspected mutation.

[0105] wherein the median mapping quality (MMQ) indicates the median of the mapping quality value of the base in the read sequence that is the same as the reference genome base for the suspected mutation site. In some examples, the median mapping quality includes the median mapping quality value of the mutant allele (ALT) in the read sequence for the suspected mutation site. In some examples, the median mapping quality includes the median mapping quality value of the reference allele (REF) in the reference genome for the suspected mutation site. In some examples, the median mapping quality includes the median mapping quality value of the mutant allele (ALT) in the read sequence for the suspected mutation site, and the median mapping quality value of the reference allele (REF) in the reference genome for the suspected mutation site. For example, MMQ = 60, 60 indicates that the median mapping quality value of the mutant allele and the reference allele are both 60.

[0106] wherein the median mapping quality by mutant (MMQM) indicates the median of the mapping quality value of the mutant base for the suspected mutation.

[0107] wherein the median position (MPOS) indicates the median of the distance of the suspected mutation site from the end of the read sequence in which it is located. For example, MPOS = 34 indicates that the median distance of the suspected mutation site from the start of the read sequence is 34.

[0108] wherein the Negative Allele Fraction in Normal (NALOD) represents the negative log of the probability that the same mutation with the same frequency as the suspected mutation is false positive in the sequencing data of the normal sample. In some applications, the NALOD can represent the negative log of the probability that the mutation with the same frequency as the tumor is false positive in the normal sample. For example, if NALOD = 1.94, it means that the suspected mutation site is not observed in the sequencing data of the normal sample.

[0109] wherein the Normal Log Odds (NLOD) represents the log of the odds ratio that the suspected mutation site is either heterozygous or homozygous in the sequencing data of the normal sample. In some examples, the NLOD is used to measure the likelihood that the mutation is a true germline mutation in the normal sample to assist in determining whether the mutation in the tumor sample is a somatic mutation. Since germline mutations are classified as either diploid heterozygous or homozygous, either one is a germline mutation. The smaller the value, the less likely the mutation in the normal sample is a germline mutation, and the more likely the mutation in the tumor sample is a somatic mutation. A low NLOD indicates that the suspected mutation site is more likely to be a disease variant. For example, if NLOD = 47.72, it means that the suspected mutation site is very unlikely to be either heterozygous or homozygous in the sequencing data of the normal sample.

[0110] wherein the Tumor Log Odds (TLOD) represents the log of the odds ratio that the suspected mutation site is a true somatic mutation, representing the likelihood that the suspected mutation site exists in the abnormal sequence data. In some application examples, the TLOD is used to directly measure the likelihood that the mutation detected from the tumor sample is a somatic mutation. The larger the value, the more likely it is a true somatic mutation. For example, TLOD = 553.69 is a very high score, indicating that the likelihood of the site having a mutation in the disease sample is very high.

[0111] wherein the NALOD and NLOD can indirectly measure the likelihood that a mutation is a true somatic mutation, and the TLOD can directly measure the likelihood that a mutation is a true somatic mutation.

[0112] The bam-readcount software processes the gene sequencing data to obtain a readcount file, and the third mutation feature data in the readcount file can provide comprehensive sequencing and alignment information for the suspected mutation site, providing important feature basis for gene mutation detection of the suspected mutation site.

[0113] In some alternative embodiments, the third mutation feature data includes at least one of an average mapping quality value, an average base quality value, an average position ratio, an average number of mismatches ratio, and an average mismatch quality value of the suspected mutation site.

[0114] In some embodiments, the third mutation feature data includes at least one of an average mapping quality value, an average base quality value, and an average position ratio of the suspected mutation site.

[0115] In some embodiments, the third mutation feature data includes an average mapping quality value, an average base quality value, and an average position ratio of the suspected mutation site.

[0116] wherein the average mapping quality value (avg_mapping_quality, amq) is an average of mapping quality values corresponding to all detected mutation bases of the determined genetic site in the reference genome. The mapping quality value is an index for measuring the matching degree of a base in the genetic variation site to the reference genome. For example, in a human nucleic acid sample, the reference Figure 3 For a determined genetic site (base type is adenine A) of the human reference genome, a plurality of nucleic acid fragments containing the genetic variation corresponding to the genetic site are detected (such as Figure 3 The variation is shown in order as thymine T, guanine G, cytosine C, and thymine T), the mapping quality value refers to the credibility of the genetic variation (thymine T, guanine G, cytosine C, and thymine T) in each nucleic acid fragment as the genetic variation of the determined genetic site (adenine A) in the human reference genome. The mapping quality value usually takes a value range of 0-40, and the higher the mapping quality value, the more reliable the matching result of the base to the determined genetic site of the human reference genome.

[0117] wherein the average base quality value (avg_base_quality, ab) is an average of base quality values of each base of the suspected mutation site identified in each read sequence. The base quality value is an index for measuring the sequencing accuracy of a single base, usually represented by Q, which can be calculated by the formula Q = -10 x log10(E), wherein E is the base identification error rate, which can be derived from the sequencing platform. The higher the base quality value, the higher the sequencing accuracy of the base.

[0118] The average position fraction (avg_pos_as_fraction, apf) is an average value of a position fraction of a base position of a suspected mutation site in each read sequence containing the suspected mutation site relative to a reference base position of a nucleic acid fragment. For a nucleic acid molecule to be sequenced, as the number of sequencing rounds increases, the sequencing performance gradually decreases when the number of sequencing rounds reaches a certain number. Therefore, the average position fraction can be used to reflect the relative position change of the suspected mutation site in different read sequences. The reference base position can be set artificially. For example, the reference base position is the position of the central base in the read sequence, i.e., the center position. For example, the position fraction of the center position of the read sequence is assigned as 1 (the variation event occurs in the middle of the read sequence), and the position fraction of the end position of the read sequence will tend to 0, because the variation event occurs at the end of the read sequence, and the sequencing performance of the base relative to the front position is poor. The average position fraction helps to identify bias patterns in sequencing data, for example, if most genetic variations are concentrated at the end of the read sequence, which may indicate a problem in the sequencing or analysis process.

[0119] The average number of mismatches fraction (avg_num_mismatches_as_fraction, anmf) is an average value of a number fraction of bases different from the reference genome (or said suspected mutation site) in each read sequence compared with the reference genome, which can also be referred to as an average value of the number of mismatches. The number fraction is a ratio between the number of bases in the read sequence that do not match the reference genome and the total number of all bases in the read sequence, which is used to represent the degree of mismatch between the genetic sequencing data and the reference genome, wherein the total number of all bases can also be understood as the length of the read sequence. Specifically, the higher the average number of mismatches fraction, the higher the degree of mismatch of the genetic variation site.

[0120] The average sum of mismatch qualities (avg_sum_mismatch_qualities, asmq) is an average value of base quality values of bases different from the reference genome in each read sequence corresponding to the suspected mutation site identified compared with the reference genome. The mismatch quality value is an index for measuring the quality of a single mismatch base, and the higher the total sum corresponding to the average mismatch quality value, the higher the uncertainty of the suspected mutation site.

[0121] The bam-readcount software processes the genetic sequencing data to obtain a readcount file, and the third mutation feature data in the readcount file can provide comprehensive sequencing and alignment information of the suspected mutation site, and provide an important feature basis for genetic mutation detection of the suspected mutation site.

[0122] S230, input the second mutation feature data and the third mutation feature data into the pre-trained target mutation detection model to obtain an output target mutation detection result of a suspected mutation site.

[0123] S230 in this embodiment corresponds to S140 shown in the above embodiments, and the details are not repeated here. Figure 1 S230 in this embodiment corresponds to S140 shown in the above embodiments, and the details are not repeated here.

[0124] The technical solution of this embodiment sets the module architecture of the first mutation detection module as Varscan software, and the module architecture of the second mutation detection module as Mutect2 software or Strelka2 software, wherein the Varscan software satisfies the constraint condition that the recall rate is greater than or equal to 0.9, and the Mutect2 software satisfies the constraint condition that the precision rate is greater than or equal to 0.9, thereby further improving the precision rate and recall rate of the target mutation detection model.

[0125] On the basis of the above embodiments, the target mutation detection model in the embodiments of the present application is obtained by pre-training. In some embodiments, the training method of the target mutation detection model comprises:

[0126] 1) obtaining training mutation sites of a training nucleic acid sample through the first mutation detection module.

[0127] In this step 1), the training nucleic acid sample is a nucleic acid sample with known genetic mutation sites, for example, a standard nucleic acid sample. The case of known genetic mutation sites includes that the nucleic acid sequence of the training nucleic acid sample is known, so that the genetic mutation site can be determined through the known sequence; or the position of the genetic mutation site in the nucleic acid sequence and the mutation type can be determined. For the training nucleic acid sample, it can be a nucleic acid sample with one known genetic mutation site, or a nucleic acid sample with two or more known genetic mutation sites.

[0128] Generally, increasing the number of training nucleic acid samples is conducive to improving the accuracy of the model when training the initial mutation detection model. Based on the fact that nucleic acid samples with known genetic mutation sites are difficult to collect in actual life, in some implementation cases, nucleic acid samples with genetic mutation sites can be generated by simulating standard samples, i.e., training nucleic acid samples.

[0129] As an embodiment, the training nucleic acid sample provided by the embodiments of the present application is a simulated tumor standard sample, and the corresponding training sequencing data is sequencing data of the simulated tumor standard sample, i.e., simulated tumor standard sample sequencing data. The simulated tumor standard sample sequencing data referred to in the embodiments of the present application is a set of data artificially constructed by using a calculation method. This set of data simulates the sequencing results of an actual tumor sample. The simulated tumor standard sample sequencing data is based on modification or processing of sequencing data of an existing cell or tissue to introduce specific variation or mutation characteristics, thereby simulating the genomic characteristics of a tumor cell. It should be noted that the simulated tumor standard sample sequencing data is not directly derived from sequencing of a real tumor sample, but is synthesized by researchers according to known tumor genomic characteristics and the characteristics of sequencing technology through algorithm design. These data contain simulated somatic mutations and other tumor-related genomic characteristics. This simulated data can be used to evaluate and optimize sequencing technology and bioinformatics analysis algorithms and processes, as well as to train and test machine learning models.

[0130] In one implementation scenario, the method for constructing simulated tumor standard sample sequencing data comprises:

[0131] Step (a), for a target region, a first set of germline mutation sites from a first human genome standard sample and a second set of germline mutation sites from a second human genome standard sample are obtained respectively.

[0132] In this step, by obtaining sets of germline mutation sites from different human genome standard samples, a basis is provided for subsequent construction of simulated tumor standard sample sequencing data.

[0133] In this step, the first set of germline mutation sites and the second set of germline mutation sites are respectively from different human genome standard samples, i.e., in order to construct simulated tumor standard sample sequencing data, germline mutation sites need to be obtained from two different human genome standard samples. These standard samples are known reference samples for scientific research and quality control, and usually contain specific genetic characteristics, such as germline mutations. According to the embodiments of the present application, these standard samples can have different germline mutations, so that the combination of two different standard samples can introduce simulated somatic mutations.

[0134] According to the embodiments of the present application, by using sets of germline mutation site data from different standard samples, the genomic characteristics of the tumor standard sample can be more accurately simulated, the performance of the sequencing platform in detecting somatic mutations can be evaluated, and positive (mutation sites) and negative (sites that have not undergone mutations) can be clearly distinguished, thereby providing a means for comprehensively evaluating the performance of sequencing technology and bioinformatics algorithms and processes.

[0135] According to embodiments of the present application, these standards can come from authorities such as the Genome in a Bottle Consortium (GIAB) or other biomedical research organizations that provide standard reference materials for genomic studies. In some examples of the present application, the aforementioned first human genome standard and the aforementioned second human genome standard are each independently selected from at least one of HG001, HG002, HG003, HG004, and HG005, and the aforementioned first human genome standard and the aforementioned second human genome standard are different. The first set of germline mutation sites and the second set of germline mutation sites obtained using these standards can be used to create simulated tumor standards, which can further help to evaluate and optimize the performance of the analysis pipeline, including their sensitivity, specificity, and accuracy.

[0136] According to embodiments of the present application, in order to construct simulated tumor standards to obtain their sequencing data, it is not necessary to use all germline mutation sites of the whole genome as the set of germline mutation sites, but only the germline mutation sites associated with the region of interest (i.e., the "targeted region") can be used as the basis for constructing the sequencing data of the simulated tumor standards. In this context, the term "targeted region" refers to a portion of the genome that is of particular interest in genomic sequencing or other molecular analysis. According to embodiments of the present application, these regions are selected for further study due to their biological significance, potential links to disease, or the inclusion of specific functional genes. According to embodiments of the present application, the targeted region includes a specific contiguous DNA sequence in the genome that is pre-selected, such as one or more genes, a portion of a gene, a regulatory region, or other genetic markers. According to embodiments of the present application, these regions can contain genes associated with a particular biological function or disease, such as a susceptibility gene for a certain cancer. It should be noted that the targeted region can be a small portion of the whole genome, or it can be multiple dispersed regions, depending on the specific needs of the study. In some examples of the present application, the aforementioned targeted region is determined based on the aforementioned specific hypermutation regions of tumor somatic mutations. The germline mutation sites obtained based on the targeted region are a set of targeted germline mutation sites. In this context, the term "hypermutation region" refers to specific genomic fragments in tumor somatic cells that have a higher mutation frequency in tumor somatic cells and are associated with the occurrence, development, or treatment response of tumors.

[0137] In some examples of the present application, the aforementioned germline mutations such as SNVs or InDels are selected from high confidence intervals to cover more mutation scenarios.

[0138] Germline mutations refer to mutations that occur in germ cells, which can be inherited by offspring. Unlike somatic mutations, germline mutations exist in all cells of an individual. Generally speaking, germline mutations can affect the mechanisms of tumorigenesis, including processes such as cell growth, differentiation, and apoptosis. According to embodiments of the present application, in tumor sample sequencing, distinguishing germline mutations from somatic mutations is crucial to reducing false positive results. By studying germline mutations, the accuracy of mutation detection can be improved. In addition, using germline mutations can conduct genomic research on diseases without involving patient privacy and ethical issues. In embodiments of the present application, the "high confidence interval" is selected from the high confidence interval given by an authoritative agency (such as GIAB, etc.) for a cell standard. An authoritative agency such as GIAB will give both a high confidence interval and a germline mutation for a cell standard, and the germline mutation is not necessarily in the high confidence interval. If the germline mutation is within the high confidence interval, both positive sites and negative sites have a high degree of reliability. The reliability of positive sites and negative sites that are not in the high confidence interval is relatively low.

[0139] After obtaining the first set of germline mutation sites and the second set of germline mutation sites, the method for constructing simulated tumor standard sequencing data according to the present application further comprises step (b), that is: selecting a set of unique germline mutation sites from the first set of germline mutation sites with respect to the second set of germline mutation sites.

[0140] According to embodiments of the present application, the first set of germline mutation sites and the second set of germline mutation sites are respectively from two different human genome standards, and both sets contain germline mutation information. By comparing the two sets, germline mutation sites that exist in the first set of germline mutation sites but do not exist in the second set of germline mutation sites can be identified, which are the unique germline mutation sites. The purpose of selecting unique germline mutation sites is to construct simulated tumor standard sequencing data, which will be used for subsequent mutation detection performance evaluation. According to embodiments of the present application, by using unique germline mutation sites, the ability of the sequencing platform in detecting mutations, including sensitivity, specificity, and accuracy, can be evaluated.

[0141] In some examples of the present application, after obtaining the set of unique germline mutation sites, the set of unique germline mutation sites can be further preprocessed in advance. Thus, it can be ensured that the sequencing data for simulating tumor standards is as accurate and reliable as possible, avoiding the influence of false data on the analysis results.

[0142] In some examples of the present disclosure, the foregoing preprocessing comprises removing, from the set of unique germline mutation sites, low-confidence germline mutation sites that are also in the second set of germline mutation sites. In some examples of the present disclosure, the second set of germline mutation sites can contain germline mutation sites that are also in the set of unique germline mutation sites. These germline mutation sites can not be identified in determining the set of unique germline mutation sites because of their low confidence. Therefore, to simulate the specificity of the tumor standard, it is necessary to remove those unique germline mutation sites that also appear in the second set of germline mutation sites. By removing non-unique germline mutation sites, preprocessing helps to purify the set of unique germline mutation sites, which can make the simulation data closer to the research target. The unique germline mutation sites after preprocessing will more accurately reflect the genetic characteristics of the tumor sample, thereby improving the reliability of the simulation data in subsequent analysis. In some examples of the present disclosure, the foregoing preprocessing comprises removing at least one unique mutation site from the set of unique mutation sites based on the positional relationship between the unique mutation sites in the set of unique mutation sites.

[0143] In some examples of the present disclosure, at least one unique germline mutation site is removed from the set of unique germline mutations based on the distance between the unique germline mutation sites in the set of unique germline mutation sites.

[0144] In some examples of the present disclosure, the foregoing preprocessing comprises determining the distance between the unique germline mutation sites in the set of unique germline mutations; and removing at least one unique germline mutation site from the set of unique germline mutations when the distance between the unique germline mutation sites is less than a predetermined distance threshold, so that the distance between the remaining unique germline mutation sites is equal to or greater than the predetermined distance threshold.

[0145] In some examples of the present disclosure, the foregoing predetermined distance threshold is related to the sequencing read length of the sequencing platform used to obtain the sequencing data. If two unique germline mutation sites used for simulation are too close, a sequencing read produced by sequencing is likely to contain more than one unique germline mutation site used for simulation, resulting in mutual interference during simulation. Therefore, the distance threshold of the unique germline mutation sites used for simulation is not less than 250 bp. Alternatively, 250 bp, 260 bp, 270 bp, 280 bp, 290 bp, 300 bp, 310 bp, 320 bp, 330 bp, 340 bp, 350 bp, 360 bp, 370 bp, 380 bp, 390 bp, or 400 bp.

[0146] According to embodiments of the present application, the pre-processing takes into account the positional relationship between the unique germline mutation sites, which can ensure the representativeness and rationality of distribution of the unique germline mutation sites in the simulated data. If the unique germline mutation sites are too close to each other, it is possible that a read generated in the sequencing process contains multiple unique germline mutation sites at the same time, which can cause confusion and interference in the analysis. Therefore, it is necessary to remove those unique germline mutation sites that are too close to other unique germline mutation sites. Thus, in order to ensure the accuracy of simulation, a minimum distance threshold is set to ensure that there is enough space between the unique germline sites used for simulation to avoid mutual interference. This threshold can be adjusted according to the read length of the sequencing platform and the experimental design. For unique germline mutation sites with a distance not exceeding the predetermined threshold, the pre-processing will remove at least one unique germline mutation site in the set of unique germline mutations so that the distance between the remaining unique germline mutation sites is equal to or greater than the predetermined distance threshold. This helps to reduce data complexity while maintaining the representativeness of the simulated data.

[0147] According to embodiments of the present application, through these optimization steps, the quality of the simulated data can be improved to ensure the reliability and effectiveness of the simulated tumor standard sequencing data in subsequent analysis.

[0148] In embodiments of the present application, the method of constructing simulated tumor standard sequencing data further comprises step (c): obtaining sequencing data of a second human genome standard and sequencing data of a first human genome standard, and for at least one preset simulated somatic mutation site, replacing the sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the second human genome standard with the sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the first human genome standard according to a predetermined replacement ratio, so as to obtain simulated tumor standard sequencing data containing simulated somatic mutations, wherein the preset simulated somatic mutation site is selected from the set of unique germline mutation sites.

[0149] Before this step, the sets of first and second germline mutation sites have been obtained from two different human genome standards, and the set of unique mutation sites has been screened therefrom. Therefore, at least one specific unique germline mutation site can be selected as the target site of the simulated tumor standard sequencing data, which is predetermined for simulating somatic mutations in tumors.

[0150] According to an embodiment of the present application, for a preset simulated somatic mutation site, the sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the second human genome standard sample is replaced with the sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the first human genome standard sample according to a certain rule, wherein the preset simulated somatic mutation site is selected from the set of unique germline mutation sites. The purpose of this operation is to simulate the somatic mutation in the tumor standard sample with germline mutations. It should be noted that the replacement operation is not random, but is related to the mutation frequency of the aforementioned preset simulated somatic mutation site. Mutation frequency refers to the probability or proportion of a specific mutation occurring in a simulated tumor sample. The data used for replacement comes from the sequencing data of the first human genome standard sample, which means that the replacement data is also based on actual sequencing, ensuring the authenticity and reliability of the simulated data. Through the above replacement operation, the simulated tumor standard sample sequencing data is finally generated. Through the introduction of this simulated data, the lack of reliable tumor standard samples can be basically solved, and the major problem of being difficult to clearly identify negative sites based on real tumor tissue standard samples is remedied. Compared with other simulation methods, first, compared with the simulation method of mixing two normal cell standard samples at the sample level, mixing at the sequencing data level can flexibly select the required somatic mutation and accurately control its frequency, so there is no need to repeatedly sequence for different frequency simulation schemes, thereby greatly saving costs; second, compared with the simulation method of modifying bases in the sequencing data of one normal cell standard sample, mixing at the sequencing data level can always maintain the true sequencing results. In addition, using simulated data can conduct genomic research of diseases without involving patient privacy and ethical issues. Furthermore, the simulated data can also provide a controllable and standardized test environment for evaluating and optimizing sequencing technology, analysis algorithms, data processing procedures, and training and testing machine learning models. This method can be widely applied in the fields of biomedical research, drug development, clinical diagnosis, and personalized medicine, and has important value in tumor genomics research and precision medicine, in particular.

[0151] The above method compensates for the deficiency of real tumor cell sequencing in clearly identifying negative sites by clearly identifying mutant and non-mutant sites, realizes data standardization and controllability, thereby providing consistent evaluation benchmarks for different sequencing platforms, facilitating performance comparison. The method can also reduce false positive results, improve detection accuracy and reliability, and has flexibility and adjustability, which can adapt to different research needs. According to the embodiments of the present application, the method of the present application overcomes the limitations of traditional sequencing methods that cannot determine negative sites by accurately identifying mutant and non-mutant sites. This clarity provides clear guidance for researchers to help understand the genetic variation in the sample. Through standardization, the method ensures the consistency and comparability of the simulated data. This means that data generated by different laboratories or research teams can be compared and analyzed under the same standard. In addition, according to the embodiments of the present application, the frequency and type of mutations in the simulated data can be controlled, thereby simulating different tumor sample characteristics. This controllability provides great flexibility for studying specific types of tumors. The method also provides consistent evaluation benchmarks for different sequencing platforms, making it possible to compare the performance of different sequencing platforms. This is of great significance for evaluating and selecting the most suitable sequencing technology. Through precise site identification and controllable simulation process, the method significantly reduces the occurrence of false positive results, thereby improving the accuracy and reliability of detection. It should be noted that by providing more accurate simulated data, the method of the present application helps to promote the development of personalized medicine, enabling treatment plans to be customized according to the specific genetic characteristics of patients. Using simulated data can also conduct genomic research on diseases without involving patient privacy and ethical issues, which is particularly important for research that needs to handle sensitive data.

[0152] For ease of understanding, the above method of constructing simulated tumor standard sequencing data is explained in detail below through specific examples.

[0153] 1) Obtain a first set of germline mutation sites of a first human genome standard HG002 and a second set of germline mutation sites of a second human genome standard HG001.

[0154] Among them, HG001 represents a standard without somatic cell mutations, and HG002 represents a standard simulating somatic mutation sources;

[0155] 2) Obtain a set of unique germline mutation sites of the first set of germline mutation sites relative to the second set of germline mutation sites.

[0156] The chromosome number (CHROM), position (POS), reference genome base (REF) and mutant base (ALT) of each germline mutation of the two standards are intersected, and the germline mutation site not present in the intersection in HG002 is the unique germline mutation site of HG002 relative to HG001 Figure 4 dark part);

[0157] Specifically, in combination with the genome, as shown in Figure 5 HG002 has a heterozygous germline mutation at a certain site, and HG001 does not, then the site is one of the unique germline mutation sites of HG002 relative to HG001.

[0158] 3) Based on the unique germline mutation site, determine the predetermined replacement ratio of each site.

[0159] For example, if a mutation frequency of 0.05 is to be obtained, and the germline mutation of HG002 at the site is homozygous, then the replacement ratio needs to be specified as 0.05; if it is heterozygous, then the replacement ratio is 0.05*2=0.1.

[0160] Further write the chromosome number, position, and replacement ratio of the above-mentioned site into a text document for simulation. In some examples of the present application, the aforementioned simulation is performed by software, such as Posmix.

[0161] In a specific example of the present application, the aligned post-sequencing data (bam file) as somatic cell mutation, the aligned post-sequencing data (bam file) as the source of simulated somatic cell mutation, and the text document containing the chromosome number, position, and replacement ratio information are input into the Posmix software. Posmix will replace the reads of HG001 at each specified site with the reads of HG002 at a predetermined replacement ratio, thereby forming a new bam file containing simulated somatic cell mutations, i.e., obtaining sequencing data containing simulated somatic cell mutations Figure 6 ).

[0162] Use of simulated tumor standard sequencing data in evaluating the detection efficiency of sequencing platform on tumor samples

[0163] In another aspect of the present application, the present application provides the use of simulated tumor standard sequencing data in evaluating the detection efficiency of sequencing platform on tumor samples, wherein the aforementioned simulated tumor standard sequencing data is obtained according to the aforementioned method for constructing simulated tumor standard sequencing data.

[0164] In some examples of the present application, the aforementioned simulated tumor standard sequencing data is stored in a server.

[0165] In some examples of the present application, the aforementioned tumor sample detection is performed using a sequencing platform. The sequencing platform is as previously described and will not be repeated here.

[0166] In some examples of the present application, the aforementioned evaluation of the detection efficiency of the tumor sample includes evaluation of the recall rate and the accuracy rate. Evaluation of the recall rate and the accuracy rate is of great significance to the detection of tumor samples. The recall rate measures the ability of the detection method to identify all actual existing mutations, reflects the sensitivity of the detection method, and ensures as few missed mutations as possible. The accuracy rate measures the overall ability of the detection method to correctly identify mutations, reflects the reliability and accuracy of the detection results. By evaluating these two indicators, researchers can comprehensively understand the performance of the detection method, optimize the detection process, select the best sequencing platform, ensure that clinical decisions are based on high-quality data, and thus improve the effectiveness of personalized treatment.

[0167] Based on the aforementioned use of the simulated tumor standard in evaluating the detection efficiency of the sequencing platform for tumor samples, the performance of the sequencing platform can be comprehensively evaluated; the reliability and accuracy of the analysis results are ensured to guide clinical decisions; and costs and time are saved.

[0168] Method for implementing an analysis process

[0169] In another aspect of the present application, a method for implementing an analysis process is provided, which comprises: inputting simulated tumor standard sequencing data into an analysis process to obtain analysis results of the analysis process, the simulated tumor standard sequencing data being constructed according to the aforementioned method for constructing simulated tumor standard sequencing data using preset simulated somatic mutation sites; and comparing the analysis results with the preset simulated somatic mutation sites used to construct the simulated tumor standard sequencing data to determine the recall rate and the accuracy rate of the analysis process.

[0170] By inputting simulated tumor standard sequencing data constructed according to preset simulated somatic mutation sites into an analysis process and comparing the analysis results with these preset simulated somatic mutation sites, the recall rate and the accuracy rate of the analysis process can be accurately calculated. This process enables researchers to comprehensively understand and evaluate the performance of the analysis process in detecting mutations. Accurate evaluation of the recall rate and the accuracy rate can help optimize and improve the detection method, select the best sequencing platform, ensure the reliability and accuracy of the analysis results, and thus improve the effectiveness of personalized treatment. In addition, this method provides a standardized evaluation means, making the performance comparison between different analysis processes and sequencing platforms more consistent and repeatable.

[0171] The training sequencing data of the training nucleic acid sample refers to nucleic acid sequence data of the training nucleic acid sample, and the nucleic acid sequence data of the training nucleic acid sample can be full sequence data of the training nucleic acid sample, can be partial sequence data of the training nucleic acid sample containing a gene mutation site, or can be a base type of a site where a gene mutation occurs, wherein the base type of the site where the gene mutation occurs includes a base type before mutation and a base type after mutation.

[0172] In the embodiments of the present application, the first mutation detection module is used to extract features from the training sequencing data of the training nucleic acid sample, and the training mutation site is obtained based on the extracted features. That is, the training mutation site is determined by using the first mutation detection module to detect mutations in the sequencing data of the training nucleic acid sample to obtain first training feature data. For each training nucleic acid sample, the training mutation site can be one, two or more. The first mutation detection module is as described above. The training mutation site of the training nucleic acid sample is obtained by using the first mutation detection module, and reference is made to the content in the above-mentioned “obtaining a suspected mutation site of a nucleic acid sample to be detected”. In order to save space, the details are not described here. Further, the “suspected mutation site” obtained by the above-mentioned method can be further confirmed to confirm the true or false positive of the suspected mutation site.

[0173] 2) inputting the second training feature data and the third training feature data of the training mutation site into the initial mutation detection model which is not trained to obtain an output prediction mutation detection result of the training mutation site.

[0174] In step 2), the second training feature data is feature data obtained by using the second mutation detection module in the above to extract features from each training mutation site, and the third training feature data is obtained by using the sequencing data processing module to process the training sequencing data to obtain total training sequencing processing data, and screening the sequencing processing data of each training mutation site in the total training sequencing processing data to obtain the third training feature data. Details of the second mutation detection module and the sequencing data processing module are not described here.

[0175] In some embodiments, the second training feature data includes the second mutation feature data in the above, and the third training feature data includes the third mutation feature data in the above, which is not described here.

[0176] For each training mutation site, the second training feature data obtained by the second mutation detection module and the third training feature data obtained by the sequencing data processing module are obtained, and are input into the initial mutation detection model which is not trained to obtain an output prediction mutation detection result of the training mutation site. In the embodiments of the present application, the initial mutation detection model is an initial model of the target mutation detection model which is not trained.

[0177] 3) training the initial mutation detection model according to the prediction mutation detection result and the standard mutation detection result corresponding to the training mutation site to obtain a target mutation detection model.

[0178] In this step, the initial mutation detection model is trained according to the prediction mutation detection result obtained in step 2) and the standard mutation detection result corresponding to the training mutation site in step 1) to obtain a target mutation detection model. It should be understood that for the same training mutation site, the prediction mutation detection result obtained in step 2) and the standard mutation detection result corresponding to the training mutation site in step 1) have a one-to-one correspondence.

[0179] In the embodiments of the present application, the model algorithm of the target mutation detection model is not strictly limited. In some embodiments, the model algorithm of the target mutation detection model adopts a gradient boosting model. Illustratively, the model algorithm of the target mutation detection model includes but is not limited to a logistic classification model, a decision tree, a support vector machine, a random forest, an adaboost model, an XGBoost model or a deep belief network, etc., which can be customized according to actual needs.

[0180] Specifically, the initial mutation detection model is trained according to the prediction mutation detection result and the standard mutation detection result corresponding to the training mutation site to obtain a target mutation detection model, including: determining a loss function according to the prediction mutation detection result and the standard mutation detection result corresponding to the training mutation site, and adjusting the model parameters of the initial mutation detection model according to the loss function until a preset end condition is met, and the initial mutation detection model in the current iteration process is taken as the target mutation detection model.

[0181] Illustratively, the function type of the loss function includes but is not limited to a square loss function, a logarithmic loss function, an exponential loss function, a mean square error loss function, a logistic regression loss function, a Huber loss function, a cross-entropy loss function and a Kullback-Leibler divergence loss function, etc. The function type of the loss function is not limited here and can be customized according to actual needs.

[0182] On the basis of the above-mentioned embodiments, before the initial mutation detection model in the current iteration process is taken as the target mutation detection model completed for verification, the method further comprises: obtaining a verification mutation site in the verification sequencing data of the verification nucleic acid sample; obtaining second verification feature data and third verification feature data of the verification mutation site extracted by the second mutation detection module and the sequencing data processing module respectively, and inputting the second verification feature data and the third verification feature data into the initial mutation detection model in the current iteration process to obtain an output verification mutation detection result of the verification mutation site; determining the model performance of the initial mutation detection model according to the verification mutation detection result, and taking the initial mutation detection model in the current iteration process as the target mutation detection model completed for verification in the case that the model performance meets the preset performance condition.

[0183] For example, the model performance includes at least one of recall rate, precision rate, F1 score, accuracy rate and ROC curve area. The model performance is not limited here and can be customized according to actual needs.

[0184] On the basis of the above-mentioned embodiments, the method further comprises: in the case that the model performance does not meet the preset performance condition, adjusting the hyperparameters of the initial mutation detection model, and returning to the step of obtaining the training mutation site in the training sequencing data of the training nucleic acid sample.

[0185] On the basis of the above-mentioned embodiments, optionally, the module architecture of the target mutation detection model is a classification model, and before the second mutation feature data and the third mutation feature data are input into the target mutation detection model pre-trained to output the mutation detection result of each suspected mutation site, the method further comprises: screening the second mutation feature data according to the feature weights corresponding to at least two second mutation features in the second mutation feature data; and screening the third mutation feature data according to the feature weights corresponding to at least two third mutation features in the third mutation feature data. Wherein, each feature weight is determined by the target mutation detection model in the last iteration training.

[0186] In some embodiments, screening the second mutation feature data according to the feature weights corresponding to at least two second mutation features in the second mutation feature data comprises: taking the second mutation feature with a feature weight greater than a first weight threshold as the screened second mutation feature. For example, the first weight threshold can be 0.001, and the first weight threshold is not limited here and can be customized according to actual needs.

[0187] In some embodiments, the second mutation feature data is filtered according to the feature weights corresponding to the at least two second mutation features respectively, including: sorting the at least two second mutation features according to the feature weights corresponding to the at least two second mutation features respectively, and determining the filtered second mutation features from the sorting result according to the first selected number or the first selected proportion. For example, the first selected number can be 5, and the first selected proportion can be 50%. Here, the first selected number or the first selected proportion is not limited, and can be set according to actual needs.

[0188] In some embodiments, the third mutation feature data is filtered according to the feature weights corresponding to the at least two third mutation features respectively in the third mutation feature data, including: taking the third mutation features with feature weights greater than a second weight threshold as the filtered third mutation features. For example, the second weight threshold can be 0.0001. Here, the second weight threshold is not limited, and can be set according to actual needs.

[0189] In some embodiments, the third mutation feature data is filtered according to the feature weights corresponding to the at least two third mutation features respectively in the third mutation feature data, including: sorting the at least two third mutation features according to the feature weights corresponding to the at least two third mutation features respectively, and determining the filtered third mutation features from the sorting result according to the second selected number or the second selected proportion. For example, the second selected number can be 10, and the second selected proportion can be 60%. Here, the second selected number or the second selected proportion is not limited, and can be set according to actual needs.

[0190] The advantage of such a setting is that the amount of mutation feature data input into the target mutation detection model can be reduced, and the calculation amount of the target mutation detection model can be improved, thereby improving the detection efficiency of gene mutations.

[0191] The technical scheme of the present embodiment solves the problem of manual screening of the first mutation detection module and the second mutation detection module by obtaining at least two pre-trained reference mutation detection models, determining the first mutation detection module and the second mutation detection module from the reference mutation detection models according to the recall rate and the precision rate corresponding to each reference mutation detection model, and improving the screening efficiency of the first mutation detection module and the second mutation detection module under the condition that the first mutation detection module and the second mutation detection module meet the performance constraint condition.

[0192] The following will be described in conjunction with specific examples.

[0193] 1) Simulated somatic mutation analysis using 3 standards of 2 sequencing platforms

[0194] Using HG001, HG002 and HG003, a total of 3 standards, there are 6 sets of sites, each set has about 300 sites. Using the tumor panel of Agena, each standard has 3 repeated sequencing data, and each repeated data has >160M reads. When simulating, the amount of tumor data used is set to 4: 50M, 70M, 100M, 130M, 150M.

[0195] 2) Use the mutation detection results of VarScan software to run Mutect2 software.

[0196] 3) Read the second training feature data extracted by Mutect2 software, and extract the third training feature data from the readcount file to form structured data.

[0197] 4) Label the real mutation sites and false positive mutation sites.

[0198] Table 1 is a structured data provided by an embodiment of the present application.

[0199]

[0200] For example, taking the training mutation site "10:104352334:C:T" in Table 1 as an example, "10" represents the chromosome number, "104352334" represents the gene mutation position, "C" represents the base at the gene mutation position in the training mutation site, and "T" represents the base at the gene mutation position in the reference genome. Among them, MBQ1 and MBQ2 represent the median of the base quality values corresponding to the mutant allele and the reference allele, respectively, MFRL1 and MFRL2 represent the median of the fragment length corresponding to the mutant allele and the reference allele, respectively, MMQ1 and MMQ2 represent the median of the alignment quality corresponding to the mutant allele and the reference allele, respectively, and label represents the labeling result of the real mutation site and the false positive mutation site, such as "1" indicating that the training mutation site is a real mutation site, and "0" indicating that the training mutation site is a false positive mutation site.

[0201] 5) Use the xgboost model for training, and perform verification and testing.

[0202] The mutation detection result of the VarScan software is applied to running the Mutect2 software in the embodiment of the application, so as to recalculate the feature value (that is, the second mutation feature data) of each training mutation site output by the VarScan software, label the real mutation sites and the false positive mutation sites respectively, and then use the above-mentioned feature value in combination with the third mutation feature data (that is, the average base quality value, the average alignment quality value and the like) extracted from the readcount file to train the xgboost model, so that the xgboost model can classify the mutation detection result of the VarScan software again to obtain the real mutation site.

[0203] Since the Mutect2 software constructs a series of feature values that can accurately distinguish the real mutation sites from the false positive mutation sites when the mutation frequency is greater than 1.5%, although a high recall rate cannot be obtained when the mutation frequency is less than 1.5%, the combination of these feature values and the third mutation feature data in the readcount file and the use of the machine learning method can train a powerful learning model for distinguishing the real mutation sites from the false positive mutation sites.

[0204] Figure 7 FIG. 1 is a schematic diagram of an xgboost software training method provided by an embodiment of the application, Figure 7 Taking the module architecture of the first mutation detection module as the Varscan software and the module architecture of the second mutation detection module as the Mutect2 software as an example. Specifically, the training mutation sites in the training sequencing data are determined through the Varscan software. The second training feature data of the training mutation sites is extracted through the Mutect2 software, the third training feature data of the sequencing data processing module in the readcount file is determined according to the second training feature data, the structured data is constituted, the xgboost model is supervised and trained according to the structured data, and after the supervision and training are completed, the performance of the xgboost model in the current iteration process is tested according to the verification sequencing data of the verification nucleic acid sample, and in the case that the performance test is passed, the xgboost model in the current iteration process is taken as the trained xgboost model.

[0205] Table 2 is model parameter data of an xgboost model provided by an embodiment of the application.

[0206]

[0207] In the table, “Feature” represents the input mutation feature, “Gain” represents the feature weight, “Cover” represents the coverage, and “Frequency” represents the frequency.

[0208] The xgboost model trained in the above specific examples is used to filter the somatic mutation detection results output by the VarScan software. At present, 0.5 is tentatively used as the cutting point of the predicted value. The following three examples are given:

[0209] 1. Filtering the simulated somatic mutation detection results of the standard samples with HG004 as the source of somatic mutations and HG005 as the standard sample without somatic mutations, and the sequencing platforms of GenoLab M and FASTASeq 300, the recall rate of the final xgboost model is 0.93, and the accuracy rate is 0.88.

[0210] 2. Filtering the simulated somatic mutation detection results of the standard samples with HG002 as the source of somatic mutations and HG001 as the standard sample without somatic mutations, and the sequencing platforms of NovaSeq 6000 and Element AVITI, the recall rate of the final xgboost model is 0.95, and the accuracy rate is 0.95.

[0211] The following is an embodiment of the gene mutation detection device provided by the present application. The device and the gene mutation detection method described above belong to the same inventive concept. Details not described in the embodiment of the gene mutation detection device can be referred to the content of the gene mutation detection method described above.

[0212] Figure 8 The structure diagram of a gene mutation detection device provided by an embodiment of the present application is shown in FIG. 1. Figure 8 As shown in the figure, the device comprises a suspected mutation site acquisition module 310, a second mutation feature data acquisition module 320, a third mutation feature data acquisition module 330, and a mutation detection result output module 340.

[0213] The suspected mutation site acquisition module 310 is configured to acquire suspected mutation sites of a nucleic acid sample to be detected.

[0214] The second mutation feature data acquisition module 320 is configured to extract features of each suspected mutation site by using a second mutation detection module to obtain second mutation feature data.

[0215] The third mutation feature data acquisition module 330 is configured to process the sequencing data by using a sequencing data processing module to obtain total sequencing processing data, and screen the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain third mutation feature data.

[0216] The mutation detection result output module 340 is configured to input the second mutation feature data and the third mutation feature data into a pre-trained target mutation detection model to output mutation detection results of each suspected mutation site.

[0217] In particular, the nucleic acid sample to be detected is used for characterizing a sample containing nucleic acid molecules or a nucleic acid mixture, and for mutation detection, wherein the nucleic acid mixture includes two or more nucleic acid molecules with different sequences. In embodiments of the present application, the term "nucleic acid molecule" refers to a polymer form of nucleotides of any length, which can be a short nucleic acid fragment of less than or equal to 500 bp in length, for example, a nucleic acid fragment of 20 bp, 50 bp, 80 bp, 100 bp, 120 bp, 150 bp, 180 bp, 200 bp, 220 bp, 250 bp, 280 bp, 300 bp, 320 bp, 350 bp, 380 bp, 400 bp, 420 bp, 450 bp, 480 bp, 500 bp, etc. in length; or a long nucleic acid fragment of greater than 500 bp in length. In embodiments of the present application, "nucleic acid molecule" includes a nucleic acid molecule composed of ribonucleotides or a nucleic acid molecule composed of deoxyribonucleotides. In addition, the nucleic acid sample to be detected in embodiments of the present application can be a nucleic acid sample after amplification, i.e., for each sequence of nucleic acid molecules in the nucleic acid sample to be detected, there are multiple copies of the nucleic acid molecules.

[0218] For example, the sample source includes, but is not limited to, blood, urine, plasma, a cell sample, an ex vivo tissue fluid or a tumor tissue, etc. The sample source of the nucleic acid sample to be detected is not limited herein. In an embodiment, the sample is a nucleic acid sample to be detected from a human, i.e., the method for detecting a genetic mutation provided in embodiments of the present application is a method for detecting a genetic mutation of a human sample. It should be understood that the method is also applicable to detecting a genetic mutation of a nucleic acid sample to be detected from other biological species. In the following description, a nucleic acid sample to be detected from a human will be taken as an example for illustration, but is not used to limit the species source of the nucleic acid sample to be detected.

[0219] In some embodiments, the nucleic acid sample to be detected is a nucleic acid sample with a mutation frequency of less than or equal to 1.5%. In particular, the genetic mutation frequency refers to the ratio of the sequencing depth of a specific mutation at a specific genomic site in a sample to the total sequencing depth. For example, in a reference genome, the total number of bases at genomic site P is 100 (i.e., the sequencing depth is 100) and the base is adenine A (i.e., if no mutation occurs, theoretically all the bases measured at genomic site P are adenine A). In the nucleic acid sample to be detected, a cytosine C mutation occurs at genomic site P and the number of cytosine C bases is 5 (i.e., the sequencing depth is 5), then the genetic mutation frequency of the nucleic acid sample to be detected is 5%.

[0220] The conventional gene mutation detection software has poor detection performance for nucleic acid samples with a mutation frequency less than or equal to 1.5%, and cannot guarantee high accuracy and high recall rate at the same time. The target mutation detection model in the embodiment is especially suitable for mutation detection of nucleic acid samples with a mutation frequency less than or equal to 1.5%, and can simultaneously consider high accuracy and high recall rate of the detection result.

[0221] In some embodiments, the nucleic acid sample to be detected is a biological sample with somatic mutations, or the nucleic acid sample to be detected is a biological sample with germ cell mutations. Among them, somatic mutations refer to mutations that occur in somatic cells during individual development due to the influence of mutagenic factors, which are usually not inherited to the next generation; germ cell mutations refer to mutations carried by germ cells during sexual reproduction, which can be inherited to the next generation.

[0222] Specifically, the sequencing data of the nucleic acid sample to be detected is used to represent the nucleic acid sequence data obtained by sequencing the nucleic acid sample to be detected using a sequencing technology. For example, the sequencing technology includes but is not limited to Sanger sequencing technology, high-throughput sequencing technology, and single-molecule sequencing technology, etc., which is not limited here and can be customized according to actual needs. The sequencing method can be single-end sequencing or double-end sequencing. Sequencing can be performed by a sequencing platform, which can be selected from but not limited to the HiSeq / MiSeq / NextSeq / NovaSeq sequencing platform of Illumina, the Ion Torrent platform of ThermoFisher / Life Technologies, the BGISEQ and MGISEQ platform of Huada, and the GenoLab M / GenoCare 1600 / FASTASeq 300 / SURFSecq5000 sequencing platform of ZMbio, without being limited thereto.

[0223] In the embodiment of the present application, the mutation site refers to a gene site in the nucleic acid sample where the sequence is inconsistent with the reference genome due to base mutation. It should be understood that the mutation site in the embodiment of the present application belongs to the category of variation sites, and the ratio of mutation sites to variation sites should be between 0 and 1.

[0224] In the embodiments of the present application, the suspected mutation site is determined by the first mutation feature data obtained by using the first mutation detection module to detect mutations in the sequencing data of the nucleic acid sample to be detected. By using the first mutation detection module to analyze the sequencing data of the nucleic acid sample to be detected, the mutation features contained in the sequencing data itself can be extracted, and / or the mutation features obtained after further data processing of the sequencing data. In the embodiments of the present application, the mutation feature data obtained by the first mutation detection module is named as the first mutation feature data, that is, the suspected mutation site in the embodiments of the present application is determined by the first mutation feature data obtained by using the first mutation detection module to detect mutations in the sequencing data of the nucleic acid sample to be detected.

[0225] In the embodiments of the present application, when the first mutation detection module for detecting mutations in the nucleic acid sample to be detected identifies the genetic mutation site, the recall rate is greater than or equal to the preset recall rate, that is, the recall rate of the genetic mutation site identified by the first mutation detection module is greater than or equal to the preset recall rate, so that the probability of identifying the real mutation site meets the preset probability requirement. Therefore, when the suspected mutation site obtained by the first mutation detection module is handed over to the second mutation feature data acquisition module 320, so that the second mutation detection module with the preset accuracy requirement is further analyzed, and the second mutation feature data extracted by the second mutation detection module is used as one of the input features of the target mutation detection model, the mutation detection result output module 340 can output the mutation detection result that can take into account the high recall rate and the high accuracy.

[0226] In some embodiments, the preset recall rate is 0.9, which means that the recall rate of the first mutation detection module for identifying the genetic mutation site, that is, the mutation detection, is greater than or equal to 0.9, thereby reducing the probability of missing the real mutation site. In some embodiments, the recall rate of the first mutation detection module for identifying the genetic mutation site is greater than or equal to 0.95.

[0227] In some embodiments, the suspected mutation site acquisition module 310 is specifically configured to: by using the alignment unit in the first mutation detection module, align the input sequencing data of the nucleic acid sample to be detected with the sequencing data of the reference genome to obtain at least one genetic variation site in the sequencing data of the nucleic acid sample to be detected that is different from the sequencing data of the reference genome; by using the feature extraction unit in the first mutation detection module, extract features from the at least one genetic variation site to obtain at least one first mutation feature data; and by using the mutation detection unit in the first mutation detection module, screen each genetic variation site according to the at least one first mutation feature data to obtain a suspected mutation site.

[0228] The advantage of such an arrangement is that, since the first mutation detection module has a high recall rate, the integrity of the identified suspected mutation sites is ensured, and some false positive mutation sites are filtered out, thereby reducing the data volume of the suspected mutation sites, and further improving the detection efficiency of gene mutations.

[0229] In this embodiment, the second mutation feature data is feature data obtained by extracting features of the suspected mutation sites using the second mutation detection module. Here, in order to distinguish from the first mutation feature data extracted by the first mutation detection module, the mutation feature data extracted by the second mutation detection module is named as the second mutation feature data.

[0230] The variation site can be used to represent a genetic position in the sequencing data that has a difference from the reference genome. For example, the variation type of the suspected mutation site includes, but is not limited to, a single nucleotide variation (SNV), an insertion-deletion mutation, a mismatch mutation, and a tandem repeat sequence, etc. The single nucleotide polymorphism emphasizes the mutation frequency of the suspected mutation site in the population, for example, the mutation frequency is greater than 1%, and the single nucleotide in the suspected mutation site is different from the corresponding position in the reference genome. The single nucleotide variation does not emphasize the mutation frequency of the suspected mutation site in the population, and the single nucleotide in the suspected mutation site is different from the corresponding position in the reference genome. The insertion-deletion mutation indicates that a new nucleic acid sequence fragment or a nucleic acid sequence fragment is deleted in the genetic mutation site, which is different from the reference genome. The mismatch mutation indicates that the base pair in the genetic mutation site is mismatched. The tandem repeat sequence indicates that one or more nucleotides are repeatedly present in the suspected mutation site.

[0231] The variation site can be obtained by using a sequencing data alignment tool. For example, the alignment tool can be a Bowtie2 tool or a BLAST tool, etc. Here, the alignment tool is not limited, and can be customized according to actual needs.

[0232] The mutation site refers to a genetic site in the nucleic acid sample, in which the sequence is inconsistent with the reference genome due to base mutation, including a single nucleotide variation and a single nucleotide variation. It should be understood that the mutation site in the embodiments of the present application belongs to the category of the variation site, and the ratio of the mutation site to the variation site should be between 0 and 1.

[0233] Here, in order to distinguish from the first mutation feature data extracted by the first mutation detection module, the mutation feature data extracted by the second mutation detection module is named as the second mutation feature data.

[0234] In the embodiments of the present application, the first mutation detection module and the second mutation detection module are both models for mutation detection of nucleic acid samples, and the mutation detection principle of the second mutation detection module is different from that of the first mutation detection module. In the embodiments of the present application, the second mutation detection module is introduced, and the mutation detection function in the second mutation detection module is used to extract features from the suspected mutation sites from the first mutation detection module, so as to improve the accuracy of the target mutation detection model trained by taking the extracted features as one of the features, so that the proportion of the real mutation sites in the suspected mutation sites identified by the first mutation detection module meets the preset proportion requirement.

[0235] In some embodiments, the accuracy of the second mutation detection module is greater than or equal to the preset accuracy, and the preset accuracy meets: the accuracy of the target mutation detection model is greater than or equal to 0.8. At this time, by setting the first mutation detection module meeting the preset recall rate requirement, the probability of the real mutation sites being identified can be improved, and then the suspected mutation sites obtained by the first mutation detection module are further analyzed by the second mutation detection module meeting the preset accuracy requirement. When the extracted second mutation features are taken as one of the input features of the target mutation detection model, the mutation detection result output by the target mutation detection model in the mutation detection result output module 340 can be a mutation detection result with high recall rate and high accuracy.

[0236] In some embodiments, the preset accuracy is 0.9, which means that the accuracy of the second mutation detection module in identifying the gene mutation site, i.e., the mutation detection, is greater than or equal to 0.9. At this time, in identifying the gene mutation site, the first mutation detection module has a recall rate greater than or equal to the preset recall rate, and the second mutation detection module has an accuracy greater than or equal to the preset accuracy. Therefore, by setting the first mutation detection module meeting the preset recall rate requirement, the probability of the real mutation sites being identified can be improved, and then the suspected mutation sites obtained by the first mutation detection module are further analyzed by the second mutation detection module meeting the preset accuracy requirement. When the extracted second mutation features are taken as one of the input features of the target mutation detection model, the mutation detection result output by the mutation detection result output module 340 can be a mutation detection result with high recall rate and high accuracy.

[0237] In some embodiments, the first mutation detection module has a recall rate of greater than or equal to 0.9 in identifying the genetic mutation site, and the second mutation detection module has a precision rate of greater than or equal to 0.9 in identifying the genetic mutation site. Thus, the probability of a true mutation site being missed is reduced by the first mutation detection module, thereby increasing the probability of a suspected mutation site being identified. The suspected mutation site obtained by the first mutation detection module is further analyzed by the second mutation detection module to reduce the probability of a false positive mutation site being missed. Finally, the second mutation feature data obtained by the second mutation detection module is used as one of the input features of the target mutation detection model, which can make the mutation detection result output by the target mutation detection model in the mutation detection result output module 340 have high recall rate and high precision rate.

[0238] In the embodiments of the present application, the recall rate and the precision rate can be obtained by testing the first mutation detection module and the second mutation detection module using a test set. The test set includes test sequencing data of test nucleic acid samples.

[0239] In one embodiment, the first mutation detection module and the second mutation detection module can be obtained by screening from the currently used genetic mutation detection model. For example, the module architecture of the first mutation detection module is Varscan software, and the module architecture of the second mutation detection module is Mutect2 software. At this time, in identifying the genetic mutation site, the recall rate of the Varscan software is high, which can reach 0.96, thereby increasing the probability of a true mutation site being identified. The suspected mutation site obtained by the Varscan software is further analyzed by the Mutect2 software with a precision rate of greater than or equal to 0.9, and the second mutation feature data obtained by the Mutect2 software is used together with the third mutation feature data extracted from the sequencing data for the suspected mutation site as one of the input features of the target mutation detection model, which can make the mutation detection result output by the target mutation detection model 340 have high recall rate and high precision rate. The results show that the recall rate and the precision rate of the target mutation detection model in detecting the mutation site are both optimal.

[0240] In some embodiments, the third mutation feature data acquisition module 330 is specifically configured to: for each suspected mutation site, obtain the sequencing processing data corresponding to the suspected mutation site in the total sequencing processing data, and add the sequencing processing data to the third mutation feature data.

[0241] In some embodiments, the third mutation feature data acquisition module 330 is specifically configured to: for each suspected mutation site, obtain the sequencing processing data corresponding to the suspected mutation site in the total sequencing processing data; and perform feature screening on the sequencing processing data, and add the screened sequencing processing data to the third mutation feature data.

[0242] In the embodiments of the present application, the representation form of the mutation detection result can be whether it is a positive site, a level, a score or a probability, etc. The level can be a true mutation level or a false positive mutation level, the score can be a true mutation score or a false positive mutation score, and the probability can be a true mutation rate or a false positive mutation rate, etc. Taking the true mutation level, the true mutation score or the true mutation rate as an example, the higher the true mutation level, the true mutation score or the true mutation rate, the greater the possibility that the suspected mutation site is a true mutation site, and vice versa, the lower the true mutation level, the true mutation score or the true mutation rate, the smaller the possibility that the suspected mutation site is a true mutation site.

[0243] Here, the representation form of the target mutation detection result is not limited and can be defined and set according to actual needs.

[0244] In an optional embodiment, the mutation detection result is a result of the suspected mutation site being a positive site.

[0245] In an optional embodiment, the mutation detection result is a result of the suspected mutation site being a negative site.

[0246] The result of the positive site can be confirmed by the true mutation level, the true mutation score and the true mutation rate, but the index for confirming the suspected mutation site as a positive site is not limited thereto. The result of the negative site can be confirmed by the false positive mutation level, the false positive mutation score and the false positive mutation rate, and similarly, the index for confirming the suspected mutation site as a negative site is not limited thereto.

[0247] In some embodiments, the second mutation feature data and the third mutation feature data are input into a pre-trained target mutation detection model, and a mutation detection result of each suspected mutation site is output, including:

[0248] The second mutation feature data and the third mutation feature data are input into a pre-trained target mutation detection model to obtain a predicted value of the suspected mutation site being positive;

[0249] The predicted value is compared with a preset target value, and a comparison result is output:

[0250] If the predicted value is greater than the preset target value, a result of the suspected mutation site being a positive site is output;

[0251] If the predicted value is less than the preset target value, a result of the suspected mutation site being a negative site is output.

[0252] The technical scheme of the embodiment, by adopting the first mutation detection module with a recall rate greater than or equal to a preset recall rate to obtain the suspected mutation site of the nucleic acid sample to be detected, ensures the integrity of the suspected mutation site detected; on this basis, the second mutation detection module is used to extract the second mutation feature data from the suspected mutation site, and the sequencing data processing module is used to process the sequencing data of the suspected mutation site, obtain the total sequencing processing data, and screen the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain the third mutation feature data; the second mutation feature data and the third mutation feature data are input into the target mutation detection model pre-trained, and the mutation detection result of each suspected mutation site is output. The target mutation detection model provided by the application is based on the first mutation feature data with high recall rate contribution, and combines the second mutation feature data extracted by the second mutation detection module different from the first mutation feature detection module, and the third mutation feature data associated with the suspected mutation feature screened out in the sequencing data. The combination of the three mutation features solves the problem of poor detection performance of the traditional mutation detection method or software system, and balances the high accuracy and high recall rate of gene mutation detection.

[0253] In some embodiments, the module architecture of the first mutation detection module is Varscan software, the module architecture of the second mutation detection module is Mutect2 software or Strelka2 software, and the sequencing data processing module is bam-readcount software.

[0254] In the embodiment, the nucleic acid sample to be detected is a nucleic acid sample to be detected from somatic cells, and the suspected mutation site is determined by the first mutation feature data obtained by using Varscan software to detect mutations in the sequencing data of the nucleic acid sample to be detected.

[0255] VarScan software is a software for somatic mutation detection under a java linux system. With the increase of sequencing depth, the recall rate of VarScan software can reach about 0.96, but the accuracy of VarScan software is poor. In the prior art, after mutation detection by VarScan software, methods such as genome annotation, genetic annotation and manual judgment are still needed to further verify the mutation detection results output by VarScan software.

[0256] The second mutation feature data is feature data obtained by extracting features of suspected mutation sites determined by VarScan software using Mutect2 software or Strelka2 software. The Mutect2 software or Strelka2 software is software for detecting somatic mutations. The Mutect2 software is based on a Bayesian algorithm of a hidden Markov model and mainly detects single nucleotide site variations and insertion-deletion variations of somatic cells. The Strelka2 software rearranges gene variation sites of insertion-deletion variations, and detects mutations based on a Bayesian probability model and mutation feature data of the rearranged gene variation sites. With the increase of sequencing depth, the recall rate of the Mutect2 software or the Strelka2 software reaches a platform value. The platform value is affected by the average error rate of the sequencing platform to which the sequencing technology belongs, so the platform value is not very high, but the precision of the Mutect2 software or the Strelka2 software can reach more than 0.9.

[0257] In some embodiments, the second mutation feature data includes at least one of sequencing depth, number of variant events, germline variant quality score, median reference base quality value, median mutant base quality value, median reference base insertion length, median mutant base insertion length, median reference base alignment quality value, median mutant base alignment quality value, distance median, no mutation sample false positive log odds, no mutation sample log likelihood ratio, and disease log likelihood ratio.

[0258] In some embodiments, the second mutation feature data includes at least one of sequencing depth, number of variant events, germline variant quality score, median reference base quality value, median mutant base quality value, median reference base insertion length, median mutant base insertion length, median reference base alignment quality value, median mutant base alignment quality value, distance median, no mutation sample false positive log odds, no mutation sample log likelihood ratio, and disease log likelihood ratio.

[0259] In some embodiments, the second mutation feature data includes at least one of sequencing depth, number of variant events, germline variant quality score, median reference base quality value, median mutant base quality value, median reference base insertion length, median mutant base insertion length, median reference base alignment quality value, median mutant base alignment quality value, distance median, no mutation sample false positive log odds, no mutation sample log likelihood ratio, and disease log likelihood ratio.

[0260] The interpretation of each second mutation feature data is as described above, and will not be repeated here for the sake of brevity.

[0261] The bam-readcount software processes the gene sequencing data to obtain a readcount file. Third mutation feature data in the readcount file can provide comprehensive sequencing and alignment information of the suspected mutation site, and provide an important feature basis for gene mutation detection of the suspected mutation site.

[0262] In some optional embodiments, the third mutation feature data includes at least one of an average alignment quality value, an average base quality value, an average position ratio, an average number of mismatches ratio, and an average mismatch quality value of the suspected mutation site.

[0263] In some embodiments, the third mutation feature data includes at least one of an average alignment quality value, an average base quality value, and an average position ratio of the suspected mutation site.

[0264] In some embodiments, the third mutation feature data includes an average alignment quality value, an average base quality value, and an average position ratio of the suspected mutation site.

[0265] The interpretation of each third mutation feature data is as described above, and will not be repeated here for the sake of brevity.

[0266] The technical scheme of the embodiment sets the module architecture of the first mutation detection module as Varscan software, and the module architecture of the second mutation detection module as Mutect2 software or Strelka2 software. The Varscan software satisfies a constraint condition that the recall rate is greater than or equal to 0.9, and the Mutect2 software satisfies a constraint condition that the precision rate is greater than or equal to 0.9, thereby improving the precision rate and the recall rate of the target mutation detection model.

[0267] On the basis of the above-mentioned embodiments, the target mutation detection model in the embodiment of the application is obtained by pre-training. In some embodiments, the device further includes a target mutation detection model training module, which includes:

[0268] The training mutation site acquisition unit is configured to acquire training mutation sites of a training nucleic acid sample by using the first mutation detection module.

[0269] The predicted mutation detection result output unit is configured to input the second training feature data and the third training feature data of the training mutation site into the initial mutation detection model that is not trained, to obtain an output predicted mutation detection result of the training mutation site. The second training feature data is feature data obtained by using the second mutation detection module to extract features of the training mutation site, and the third training feature data is feature data obtained by using the sequencing data processing module to process the training sequencing data.

[0270] The target mutation detection model determination unit is configured to train the initial mutation detection model according to the predicted mutation detection result and the standard mutation detection result corresponding to the training mutation site, to obtain a target mutation detection model.

[0271] In the training mutation site acquisition unit, the training nucleic acid sample is a nucleic acid sample with a known genetic mutation site, for example, a standard nucleic acid sample. The case of the known genetic mutation site includes that the nucleic acid sequence of the training nucleic acid sample is known, so that the genetic mutation site can be determined through the known sequence; or the position of the genetic mutation site in the nucleic acid sequence and the mutation type of the training nucleic acid sample can be determined. The training nucleic acid sample can be a nucleic acid sample with one known genetic mutation site, or a nucleic acid sample with two or more known genetic mutation sites.

[0272] Generally, increasing the number of training nucleic acid samples is conducive to improving the accuracy of the model when training the initial mutation detection model. However, it is difficult to collect nucleic acid samples with known genetic mutation sites in actual life. Therefore, in some implementation cases, nucleic acid samples with genetic mutation sites, i.e., training nucleic acid samples, can be generated by simulating standard samples.

[0273] As an example, the training nucleic acid sample provided by the embodiment of the present application is a simulated tumor standard sample, and the corresponding training sequencing data is the sequencing data of the simulated tumor standard sample, i.e., simulated tumor standard sample sequencing data. The simulated tumor standard sample sequencing data referred to in the embodiment of the present application is a set of data artificially constructed by using a calculation method, which simulates the sequencing results of actual tumor samples. The simulated tumor standard sample sequencing data is based on the modification or processing of existing cell or tissue sequencing data to introduce specific variation or mutation characteristics, so as to simulate the genomic characteristics of tumor cells. It should be noted that the simulated tumor standard sample sequencing data is not directly derived from the sequencing of real tumor samples, but is designed and synthesized by researchers according to known tumor genomic characteristics and the characteristics of sequencing technology through an algorithm. These data contain simulated somatic mutations and other tumor-related genomic characteristics. This simulated data can be used to evaluate and optimize sequencing technology and bioinformatics analysis algorithm and process, as well as to train and test machine learning models.

[0274] In another aspect of the present application, the device further comprises a simulated tumor standard sample sequencing data construction module, which is configured to Figure 9 The module comprises a germline mutation site set acquisition unit 100, a unique germline mutation site set acquisition unit 200, and a simulated tumor standard sample sequencing data construction unit 300. Among them,

[0275] unit 100 is configured to obtain, for each targeted region, a first set of germline mutation sites from a first human genome standard and a second set of germline mutation sites from a second human genome standard, respectively;

[0276] In some examples of the present disclosure, the targeted regions are determined based on the specific tumor somatic mutation hotspots. The germline mutation sites obtained based on the targeted regions are a set of targeted germline mutation sites.

[0277] In some examples of the present disclosure, the first human genome standard and the second human genome standard are independently selected from at least one of HG001, HG002, HG003, HG004 and HG005, and the first human genome standard and the second human genome standard are different.

[0278] In some examples of the present disclosure, the germline mutations are selected from high-confidence germline mutations, such as SNV or InDel mutations.

[0279] unit 200 is configured to select, from the first set of germline mutation sites, a set of unique germline mutation sites with respect to the second set of germline mutation sites;

[0280] In some examples of the present disclosure, after unit 200 and before unit 300, further comprising an optimization unit configured to pre-process the set of unique germline mutation sites in advance.

[0281] In some examples of the present disclosure, the optimization unit comprises a first optimization sub-unit configured to remove, from the set of unique mutation sites, low-confidence germline mutation sites that are identical to the second set of germline mutations.

[0282] In some examples of the present disclosure, the optimization unit further comprises a second optimization sub-unit configured to remove, from the set of unique germline mutation sites, at least one unique germline mutation site based on a positional relationship between the unique germline mutation sites in the set of unique germline mutation sites. Alternatively, at least one unique germline mutation site is removed from the set of unique germline mutations based on a distance between the unique germline mutation sites in the set of unique germline mutation sites. Specifically, the distance between the unique germline mutation sites in the set of unique germline mutations is determined; when the distance between the unique germline mutation sites is less than a predetermined distance threshold, at least one of the unique germline mutation sites is removed from the set of unique germline mutations, so that the distance between the remaining unique germline mutation sites is equal to or greater than the predetermined distance threshold.

[0283] In some examples of the present application, the aforementioned predetermined distance threshold is related to the sequencing read length of the sequencing platform used to obtain the aforementioned sequencing data. If the simulated unique germline mutation sites are too close, a sequencing read generated by sequencing is likely to contain more than one simulated unique germline mutation site, resulting in mutual interference during simulation. Therefore, the distance threshold of the simulated unique germline mutation site is not less than 250 bp. Alternatively, the distance threshold of the simulated unique germline mutation site is 250 bp, 260 bp, 270 bp, 280 bp, 290 bp, 300 bp, 310 bp, 320 bp, 330 bp, 340 bp, 350 bp, 360 bp, 370 bp, 380 bp, 390 bp or 400 bp.

[0284] The unit 300 is configured to obtain sequencing data of a second human genome standard sample and sequencing data of a first human genome standard sample, replace, for at least one preset simulated somatic mutation site, sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the second human genome standard sample with sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the first human genome standard sample according to a predetermined replacement ratio, so as to obtain simulated tumor standard sample sequencing data containing simulated somatic mutations, wherein the preset simulated somatic mutation site is selected from a set of unique germline mutation sites.

[0285] In some examples of the present application, the aforementioned predetermined replacement ratio is related to the mutation frequency of the preset simulated somatic mutation site.

[0286] In some examples of the present application, the aforementioned units are connected, and the aforementioned connection includes physical connection or network connection.

[0287] Based on the above-mentioned modules, simulated tumor standard sample sequencing data can be efficiently and accurately generated.

[0288] The training sequencing data of the training nucleic acid sample refers to the nucleic acid sequence data of the training nucleic acid sample. The nucleic acid sequence data of the training nucleic acid sample can be the full sequence data of the training nucleic acid sample, or the partial sequence data of the training nucleic acid sample containing a gene mutation site, or even the base type of the gene mutation site, wherein the base type of the gene mutation site includes the base type before mutation and the base type after mutation.

[0289] In the embodiments of the present application, the first mutation detection module is used to extract features from the training sequencing data of the training nucleic acid sample, and the training mutation site is obtained based on the extracted features. For each training nucleic acid sample, the training software site can be one, two or more. The first mutation detection module is as described above. The training mutation site of the training nucleic acid sample is obtained by the first mutation detection module. For brevity, the details are not repeated here. Further, the “suspected mutation site” obtained by the above method can be further confirmed to confirm the true or false positive of the suspected mutation site.

[0290] In the prediction mutation detection result output unit, the second training feature data is the feature data obtained by extracting features from the training mutation site by the second mutation detection module described above, and the third training feature data is obtained by the following method: processing the training sequencing data by the sequencing data processing module to obtain total training sequencing processing data; screening the sequencing processing data of each training mutation site in the total training sequencing processing data to obtain the third training feature data. The second mutation detection module and the sequencing data processing module are not described here.

[0291] In some embodiments, the second training feature data includes the second mutation feature data described above, and the third training feature data includes the third mutation feature data described above, which will not be described here.

[0292] For each training mutation site, the second training feature data obtained by the second mutation detection module and the third training feature data obtained by the sequencing data processing module are input into the initial mutation detection model which is not trained to obtain the prediction mutation detection result of the output training mutation site. In the embodiments of the present application, the initial mutation detection model is an initial model of the target mutation detection model which is not trained.

[0293] In the target mutation detection model determination unit, the initial mutation detection model is trained based on the prediction mutation detection result obtained by the prediction mutation detection result output unit and the standard mutation detection result corresponding to the training mutation site obtained by the training mutation site acquisition unit to obtain the target mutation detection model which is trained. It should be understood that for the same training mutation site, the prediction mutation detection result obtained by the prediction mutation detection result output unit and the standard mutation detection result corresponding to the training mutation site obtained by the training mutation site acquisition unit have a one-to-one correspondence.

[0294] In the embodiments of the present application, the model algorithm of the target mutation detection model is not strictly limited. In some embodiments, the model algorithm of the target mutation detection model adopts a gradient boosting model. Illustratively, the model algorithm of the target mutation detection model includes, but is not limited to, a logistic classification model, a decision tree, a support vector machine, a random forest, an adaboost model, an XGBoost model, or a deep belief network, and the like, which can be customized according to actual needs.

[0295] Specifically, the target mutation detection model determining unit is specifically configured to: determine a loss function according to the predicted mutation detection result and the standard mutation detection result corresponding to the training mutation site, and adjust the model parameters of the initial mutation detection model according to the loss function until a preset ending condition is met, and then take the initial mutation detection model in the current iteration process as the target mutation detection model trained.

[0296] Illustratively, the function type of the loss function includes, but is not limited to, a square loss function, a logarithmic loss function, an exponential loss function, a mean square error loss function, a logistic regression loss function, a Huber loss function, a cross-entropy loss function, and a Kullback-Leibler divergence loss function, and the like. The function type of the loss function is not limited here and can be customized according to actual needs.

[0297] On the basis of the above-mentioned embodiments, the device further includes a target mutation detection model verification module, configured to: before taking the initial mutation detection model in the current iteration process as the target mutation detection model verified, obtain a verification mutation site in verification sequencing data of a verification nucleic acid sample; obtain second verification feature data and third verification feature data of the verification mutation site extracted by the second mutation detection module and the sequencing data processing module respectively, and input the second verification feature data and the third verification feature data into the initial mutation detection model in the current iteration process to obtain a verification mutation detection result of the output verification mutation site; determine the model performance of the initial mutation detection model according to the verification mutation detection result, and take the initial mutation detection model in the current iteration process as the target mutation detection model verified in the case that the model performance meets a preset performance condition.

[0298] Illustratively, the model performance includes, but is not limited to, at least one of recall rate, precision rate, F1 score, correct rate, and area of ROC curve. The model performance is not limited here and can be customized according to actual needs.

[0299] On the basis of the above-mentioned embodiments, the device further includes a hyperparameter adjustment module, configured to: in the case that the model performance does not meet the preset performance condition, adjust the hyperparameters of the initial mutation detection model, and return to execute the step of obtaining the training mutation site of the training nucleic acid sample by the first mutation detection module.

[0300] In some embodiments, the model architecture of the target mutation detection model is a classification model; the apparatus further comprises:

[0301] The second mutation feature data screening module is configured to, before inputting the second mutation feature data and the third mutation feature data into the target mutation detection model which has been pre-trained, and outputting the mutation detection result of each suspected mutation site, screen the second mutation feature data according to the feature weights corresponding to the at least two second mutation features respectively.

[0302] The third mutation feature data screening module is configured to screen the third mutation feature data according to the feature weights corresponding to the at least two third mutation features respectively; wherein each feature weight is determined by the target mutation detection model in the last iteration training.

[0303] In some embodiments, the second mutation feature data screening module is specifically configured to: take the second mutation feature with a feature weight greater than a first weight threshold as the screened second mutation feature. Exemplarily, the first weight threshold can be 0.001, and here the first weight threshold is not limited, and can be set as needed.

[0304] In some embodiments, the second mutation feature data screening module is specifically configured to: sort the at least two second mutation features according to the feature weights corresponding to the at least two second mutation features respectively, and determine the screened second mutation feature from the sorting result according to a first selection quantity or a first selection ratio. Exemplarily, the first selection quantity can be 5, and the first selection ratio can be 50%, and here the first selection quantity or the first selection ratio is not limited, and can be set as needed.

[0305] In some embodiments, the third mutation feature data screening module is specifically configured to: take the third mutation feature with a feature weight greater than a second weight threshold as the screened third mutation feature. Exemplarily, the second weight threshold can be 0.0001, and here the second weight threshold is not limited, and can be set as needed.

[0306] In some embodiments, the third mutation feature data screening module is specifically configured to: sort the at least two third mutation features according to the feature weights corresponding to the at least two third mutation features respectively, and determine the screened third mutation feature from the sorting result according to a second selection quantity or a second selection ratio. Exemplarily, the second selection quantity can be 10, and the second selection ratio can be 60%, and here the second selection quantity or the second selection ratio is not limited, and can be set as needed.

[0307] The advantage of the arrangement is that the amount of data of the mutation features input into the target mutation detection model can be reduced, and the calculation amount of the target mutation detection model can be improved, thereby improving the detection efficiency of the gene mutation.

[0308] The technical scheme of the embodiment can obtain at least two reference mutation detection models that are pre-trained, determine the first mutation detection module and the second mutation detection module from the reference mutation detection models according to the recall rate and the precision rate corresponding to each reference mutation detection model, and solve the problem of manual screening of the first mutation detection module and the second mutation detection module, thereby improving the screening efficiency of the first mutation detection module and the second mutation detection module while ensuring that the first mutation detection module and the second mutation detection module meet the performance constraint condition.

[0309] The gene mutation detection device provided in the embodiment can execute the gene mutation detection method provided in any of the embodiments, and has the corresponding function modules and beneficial effects of the execution method.

[0310] Figure 10 A structure diagram of an electronic device is provided for an embodiment of the present application. The electronic device 10 is intended to represent various forms of digital computers, such as laptops, desktops, tablets, servers, servers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections, and relationships, and their functions, are merely examples and are not intended to limit the implementations described and / or claimed in this document.

[0311] As shown in Figure 10 The electronic device 10 includes at least one processor 11, and a memory, such as a read-only memory (Read-Only Memory, ROM) 12, a random access memory (Random Access Memory, RAM) 13, etc., which is in communication connection with the at least one processor 11, wherein the memory stores a computer program that can be executed by the at least one processor 11, and the processor 11 can execute various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (Input / Output, I / O) interface 15 is also connected to the bus 14.

[0312] A plurality of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, and the like; an output unit 17, such as various types of displays, a speaker, and the like; a storage unit 18, such as a magnetic disk, an optical disk, and the like; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 19 allows the electronic device 10 to exchange information or data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0313] The processor 11 can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various special-purpose Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, a Digital Signal Processing (DSP), and any appropriate processor, controller, microcontroller, and the like. The processor 11 performs various methods and processes described above, such as the detection method of genetic mutation provided by the above embodiments.

[0314] In some embodiments, the detection method of genetic mutation provided by the above embodiments can be implemented as a computer program tangibly embodied in a computer readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded onto the RAM 13 and executed by the processor 11, one or more steps of the detection method of genetic mutation described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to perform the detection method of genetic mutation by any other appropriate means, such as by means of firmware.

[0315] The various embodiments of the systems and techniques described above herein can be implemented in: digital electronic circuitry, integrated circuitry, a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Application Specific Standard Parts (ASSP), a System on Chip (SOC), a Complex Programmable Logic Device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0316] Computer programs used to implement the methods of detecting genetic mutations of the present application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the computer program running on the processor implements the functions / operations specified in the flowcharts and / or block diagrams. The computer program can execute entirely on a machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0317] In the context of this application, a computer-readable storage medium can be a tangible medium that can contain or store computer programs for use by or in connection with an instruction execution system, apparatus, or device. Computer-readable storage media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable storage medium. Examples of machine-readable storage media can include an electrical connection based on at least one wire, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0318] To provide for interaction with a user, the systems and techniques described here can be implemented on a terminal device having a display device (e.g., a Cathode-Ray Tube (CRT) or Liquid Crystal Display (LCD) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the terminal device. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0319] The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a Local Area Network (LAN), a Wide Area Network (WAN), a blockchain network, and the Internet.

[0320] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system, to solve the defects of large management difficulty and weak business scalability in traditional physical host and virtual private server (VPS) services.

[0321] It should be understood that the various forms of flow shown above can be reordered, added to, or have steps deleted. For example, the steps described in the present application can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions of the present application can be achieved, which are not limited herein.

[0322] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting a genetic mutation, characterized by, The method comprises: obtaining suspected mutation sites of a nucleic acid sample to be detected, wherein the suspected mutation sites are determined by first mutation feature data obtained by using a first mutation detection module to detect mutations of sequencing data of the nucleic acid sample to be detected, and the first mutation detection module has a recall rate of a gene mutation site greater than or equal to a preset recall rate; using a second mutation detection module to extract second mutation feature data from each suspected mutation site; using a sequencing data processing module to process the sequencing data to obtain total sequencing processing data, and screening sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain third mutation feature data; inputting the second mutation feature data and the third mutation feature data into a pre-trained target mutation detection model to output mutation detection results of each suspected mutation site.

2. The method of claim 1, wherein, The second mutation feature data comprises at least one of sequencing depth, number of variant events, germline variant quality score, median reference base quality value, median mutation base quality value, median reference base insertion fragment length, median mutation base insertion fragment length, median reference base alignment quality value, median mutation base alignment quality value, distance median, non-mutation sample false positive logarithmic odds, non-mutation sample logarithmic likelihood ratio, and disease logarithmic likelihood ratio. Optionally, the second mutation feature data comprises at least one of sequencing depth, number of variant events, median reference base quality value, median mutation base quality value, median reference base insertion fragment length, median mutation base insertion fragment length, median reference base alignment quality value, median mutation base alignment quality value, distance median, non-mutation sample false positive logarithmic odds, non-mutation sample logarithmic likelihood ratio, and disease logarithmic likelihood ratio. Optionally, the second mutation feature data comprises at least one of sequencing depth, number of variant events, median reference base quality value, median mutation base quality value, median reference base insertion fragment length, median mutation base insertion fragment length, median reference base alignment quality value, median mutation base alignment quality value, distance median, non-mutation sample false positive logarithmic odds, non-mutation sample logarithmic likelihood ratio, and disease logarithmic likelihood ratio. wherein the sequencing depth is used to represent the number of times a site is covered by reads; the number of variant events represents the number of variant events observed for the identified suspected mutation site; the germline variant quality score represents a quality score of the identified suspected mutation site not belonging to a germline variant, and is used to represent the possibility of the suspected mutation site not belonging to a germline variant; the median of the reference base quality values represents the median of the quality values of the bases identical to the reference genome bases corresponding to the identified suspected mutation site; the median of the mutant base quality values represents the median of the quality values of the mutant bases corresponding to the identified suspected mutation; the median of the insert length of the reference bases represents the median of the insert length of the paired reads in which the bases identical to the reference genome bases corresponding to the identified suspected mutation site are located; the median of the insert length of the mutant bases represents the median of the insert length of the paired reads in which the mutant bases corresponding to the identified suspected mutation are located; the median of the alignment quality values of the reference bases represents the median of the alignment quality values of the bases identical to the reference genome bases corresponding to the allele of the identified suspected mutation site; the median of the alignment quality values of the mutant bases represents the median of the alignment quality values of the mutant bases corresponding to the identified suspected mutation; the median of the distance represents the median of the distance of the identified suspected mutation site from the end of the read in which it is located; the log odds of false positive in no mutation sample represents the negative log of the odds that a mutation identical to and with the same frequency as the identified suspected mutation is false positive in the sequencing data of the no mutation sample; the log likelihood ratio in no mutation sample represents the log value of the likelihood ratio that the identified suspected mutation site is a true germline mutation (heterozygote or homozygote) in the sequencing data of the no mutation sample; the log likelihood ratio of disease represents the log value of the likelihood ratio that the suspected mutation site is a true somatic mutation; Optionally, the third mutation feature data comprises at least one of the average alignment quality value, the average base quality value, the average position proportion, the average proportion of number of mismatches, and the average mismatch quality value of the suspected mutation site; Optionally, the third mutation feature data comprises at least one of the average alignment quality value, the average base quality value, and the average position proportion of the suspected mutation site; Optionally, the third mutation feature data comprises at least one of the average alignment quality value, the average base quality value, and the average position proportion of the suspected mutation site; Optionally, the third mutation feature data comprises at least one of the average alignment quality value, the average base quality value, and the average position proportion of the suspected mutation site; The average alignment quality value is an average value of alignment quality values of all detected mutation bases corresponding to the determined gene site in the reference genome; the average base quality value is an average value of base quality values of each base corresponding to the identified suspected mutation site in each read sequence; the average position ratio is an average value of position ratios of base positions on the suspected mutation site relative to the reference base position of the nucleic acid fragment in each read sequence containing the same suspected mutation site; the average number of mismatches ratio is an average value of the number ratio of bases different from the reference genome in each read sequence compared with the reference genome; and the average mismatch quality value is an average value of base quality values of bases different from the human reference genome in each read sequence corresponding to the identified suspected mutation site compared with the reference genome. Optionally, the preset recall rate is 0.

9. Optionally, the recall rate of the first mutation detection module for identifying the gene mutation site is greater than or equal to 0.

95.

3. The method according to claim 1 or 2, characterized in that, The screening of the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain the third mutation feature data comprises: for each suspected mutation site, obtaining the sequencing processing data corresponding to the suspected mutation site in the total sequencing processing data, and adding the sequencing processing data to the third mutation feature data; or The screening of the sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain the third mutation feature data comprises: for each suspected mutation site, obtaining the sequencing processing data corresponding to the suspected mutation site in the total sequencing processing data; performing feature screening on the sequencing processing data, and adding the screened sequencing processing data to the third mutation feature data. Optionally, the mutation detection result is a result that the suspected mutation site is a positive site or a result that the suspected mutation site is a negative site; the inputting of the second mutation feature data and the third mutation feature data into the target mutation detection model pre-trained to output the mutation detection result of each suspected mutation site comprises: Inputting the second mutation feature data and the third mutation feature data into the target mutation detection model pre-trained to obtain a prediction value of the suspected mutation site being positive; Comparing the prediction value with a preset target value to output a comparison result: If the prediction value is greater than the preset target value, outputting a result that the suspected mutation site is a positive site; If the prediction value is less than the preset target value, outputting a result that the suspected mutation site is a negative site.

4. The method according to any one of claims 1 to 3, characterized in that, The obtaining of the suspected mutation site of the nucleic acid sample to be detected comprises: By the alignment unit in the first mutation detection module, the input sequencing data of the nucleic acid sample to be detected is aligned with the sequencing data of the reference genome to obtain at least one gene variation site different from the sequencing data of the reference genome in the sequencing data of the nucleic acid sample to be detected. The feature extraction unit in the first mutation detection module extracts features from each of the at least one genetic variant site to obtain at least one first mutation feature data; The mutation detection unit in the first mutation detection module screens each of the genetic variant sites according to the at least one first mutation feature data to obtain a suspected mutation site; Optionally, the accuracy of the second mutation detection module is greater than or equal to a preset accuracy, and the preset accuracy satisfies: the accuracy of the target mutation detection model is greater than or equal to 0.8; Optionally, the accuracy of the second mutation detection module in identifying genetic mutation sites is greater than or equal to 0.9; Optionally, the module architecture of the first mutation detection module is Varscan software, and the module architecture of the second mutation detection module is Mutect2 software.

5. The method according to any one of claims 1 to 4, characterized in that, The method further comprises: The first mutation detection module obtains training mutation sites of a training nucleic acid sample; wherein the training nucleic acid sample is a nucleic acid sample containing known mutation sites, the training mutation sites are determined by first training feature data obtained by using the first mutation detection module to detect mutations of training sequencing data of the training nucleic acid sample; the second training feature data and the third training feature data of the training mutation sites are input into an initial mutation detection model that has not been trained to obtain the predicted mutation detection result of the training mutation sites output; wherein the second training feature data is feature data obtained by using the second mutation detection module to extract features from each of the training mutation sites, and the third training feature data is obtained by: using a sequencing data processing module to process the training sequencing data to obtain total training sequencing processing data; screening sequencing processing data of each of the training mutation sites in the total training sequencing processing data to obtain the third training feature data; According to the predicted mutation detection result and the standard mutation detection result corresponding to the training mutation sites, the initial mutation detection model is trained to obtain a training completed target mutation detection model; Optionally, the training nucleic acid sample is a simulated tumor standard, and the training sequencing data is simulated tumor standard sequencing data; Optionally, the simulated tumor standard sequencing data is obtained by the following method: (a) For a target region, a first set of germline mutation sites from a first human genome standard and a second set of germline mutation sites from a second human genome standard are obtained respectively; (b) A set of unique germline mutation sites relative to the second set of germline mutation sites is selected from the first set of germline mutation sites; (c) obtaining sequencing data of the second human genome standard and sequencing data of the first human genome standard, and replacing, for at least one preset simulated somatic mutation site, sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the second human genome standard with sequencing data corresponding to the preset simulated somatic mutation site in the sequencing data of the first human genome standard according to a predetermined replacement ratio, so as to obtain the simulated tumor standard sequencing data containing simulated somatic mutations, the preset simulated somatic mutation site being selected from the set of unique germline mutation sites; Optionally, before step (c) is performed, the set of unique germline mutation sites is preprocessed; Optionally, the preprocessing comprises: removing at least one of the unique mutation sites in the set of unique mutation sites based on the positional relationship between the unique mutation sites in the set of unique mutation sites; Optionally, at least one of the unique germline mutation sites in the set of unique germline mutations is removed based on the distance between the unique germline mutation sites in the set of unique germline mutation sites; Optionally, the preprocessing comprises: determining the distance between the unique germline mutation sites in the set of unique germline mutations; when the distance between the unique germline mutation sites is less than a predetermined distance threshold, removing at least one of the unique germline mutation sites in the set of unique germline mutations so that the distance between the remaining unique germline mutation sites is equal to or greater than the predetermined distance threshold; Optionally, the predetermined distance threshold is related to the sequencing read length of the sequencing platform for obtaining the sequencing data; Optionally, the predetermined distance threshold is not less than 250 bp; Optionally, the first human genome standard and the second human genome standard are each independently selected from at least one of HG001, HG002, HG003, HG004 and HG005, and the first human genome standard and the second human genome standard are different; Optionally, the model architecture of the target mutation detection model is a classification model; Before inputting the second mutation feature data and the third mutation feature data into the target mutation detection model pre-trained, the method further comprises: screening the second mutation feature data according to the feature weights corresponding to at least two second mutation features in the second mutation feature data; screening the third mutation feature data according to the feature weights corresponding to at least two third mutation features in the third mutation feature data; wherein each of the feature weights is determined by the target mutation detection model at the last iteration training.

6. The method according to any one of claims 1 to 5, characterized in that, The nucleic acid sample to be detected is a nucleic acid sample with a mutation frequency of less than or equal to 1.5%; Optionally, the nucleic acid sample to be detected is a biological sample with somatic mutations.

7. A device for detecting a genetic mutation, comprising: comprises: The suspected mutation site acquisition module is configured to acquire suspected mutation sites of a nucleic acid sample to be detected; wherein the suspected mutation sites are determined by first mutation feature data obtained by using a first mutation detection module to detect mutations in sequencing data of the nucleic acid sample to be detected, and the first mutation detection module has a recall rate for identifying gene mutation sites greater than or equal to a preset recall rate; The second mutation feature data acquisition module is configured to use a second mutation detection module to extract features of each suspected mutation site to obtain second mutation feature data; The third mutation feature data acquisition module is configured to use a sequencing data processing module to process the sequencing data to obtain total sequencing processing data, and to screen sequencing processing data of each suspected mutation site in the total sequencing processing data to obtain third mutation feature data; The mutation detection result output module is configured to input the second mutation feature data and the third mutation feature data into a target mutation detection model that has been pre-trained, and to output mutation detection results of each suspected mutation site.

8. An electronic device, comprising: The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the method for detecting gene mutations according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to execute the method for detecting gene mutations according to any one of claims 1-6 when executed.

10. A computer program product comprising a computer program that, when executed by a processor, implements the method for detecting gene mutations according to any one of claims 1-6.

Images