Method and system for judging homologous region influence in ngs gene variant detection

Abstract

The application belongs to the technical field of gene detection, and discloses a method and system for judging the influence of homologous regions in NGS gene variation detection. The method comprises the following steps: obtaining a first type of gene and a plurality of second type of genes according to a public database; obtaining corresponding original sequence files of the first type of gene and the second type of gene based on a local gene database, obtaining corresponding sequence alignment files based on a local BAM database, and combining them to obtain known NGS gene data; calculating the GC content, the proportion of repeat units, and the similarity ratio in the original sequence file, and calculating the alignment quality in the corresponding sequence alignment file to construct a training sample; inputting the training sample into a classification model to obtain an optimized classification model; extracting the GC content, the proportion of repeat units, the similarity ratio, and the alignment quality from the NGS gene data to be detected and inputting them into the optimized classification model to judge whether the variation detection is affected. The application can judge the influence of homologous regions in variation detection without the aid of prior information of homologous regions.

Description

Technical Field

[0001] This invention relates to the field of gene detection technology, specifically to a method and system for determining the influence of homologous regions in NGS gene mutation detection. Background Technology

[0002] The NGS (Next-Generation Sequencing) platform, as the first truly high-throughput sequencing platform, has demonstrated excellent performance in solving complex biological problems related to the genome and has effectively reduced the cost of gene sequencing. Consequently, it has wide applications in precision medicine scenarios such as genetic research and clinical testing.

[0003] The basic process of gene variation detection based on the NGS platform is as follows: acquire several NGS data sets to be tested, align the NGS data with a standard reference genome, and determine the existence of gene variations based on the differences between the two sets of data. However, due to the influence of homologous regions in the reference genome, sequence alignment often cannot be performed effectively. For example, due to the presence of homologous regions, NGS data may be mapped to multiple gene fragments in the reference genome during the alignment process, resulting in the NGS data being discarded and gene sequence coverage gaps; at the same time, the alignment of homologous pseudogenes to the genome during the alignment process may also lead to the detection of false positive variations.

[0004] However, it's important to note that the sequence alignment and detection anomalies caused by homologous regions are not inevitable events, but rather related to the degree of homology between the homologous regions. In existing technologies, determining the extent to which homologous regions will affect variant detection depends on sufficient prior information about those regions. Furthermore, in actual detection processes, not all homologous regions have reliable prior information, making it impossible to determine whether the variant detection results are affected by homologous regions, and thus, to confirm the reliability of the detection results. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for judging the influence of homologous regions in NGS gene mutation detection, so as to solve the technical problem that it is impossible to assess whether homologous regions will cause detection abnormalities in the absence of prior information on homologous regions.

[0006] To achieve the above objectives, the present invention proposes the following technical solution:

[0007] Methods for determining the influence of homologous regions in NGS gene variant detection include:

[0008] Several first-class genes and several second-class genes were obtained from publicly available databases; wherein, the first-class genes are gene fragments whose mutation detection results have been confirmed to be affected by homologous regions, and the second-class genes are gene fragments whose mutation detection results have been confirmed to be unaffected by homologous regions.

[0009] Several original sequence files corresponding to the first type of gene and the second type of gene are obtained based on the local gene database. Sequence alignment files corresponding to each of the original sequence files are obtained based on the local BAM database. The historical test data and the corresponding sequence alignment files are combined to obtain several known NGS gene data.

[0010] The GC content, repetitive unit ratio, and similarity ratio in the original sequence file are calculated, and the alignment quality in the corresponding sequence alignment file is calculated; then, training samples corresponding to each of the known NGS gene data are constructed using these parameters.

[0011] With the goal of making the accuracy, precision, recall, F1 score and AUC value of the model evaluation parameters all approach 1, the training samples are input into the classification model to iteratively train it until the preset number of iterations is reached, so as to obtain an optimized classification model.

[0012] The GC content, repetitive unit ratio, similarity ratio, and alignment quality are extracted from the NGS gene data to be detected and input into the optimized classification model to determine whether the mutation detection of the corresponding gene is affected by homologous regions.

[0013] Furthermore, the step of inputting the training samples into the classification model for iterative training until a preset number of iterations is reached to obtain an optimized classification model includes:

[0014] The training samples are simultaneously input into several different types of classification models to iteratively train them until a preset number of iterations is reached, so as to obtain several different types of intermediate classification models.

[0015] The intermediate classification model with the AUC value closest to 1 is selected as the optimized classification model.

[0016] Furthermore, including:

[0017] The various types of classification models include: logistic regression model, support vector machine model, decision tree model, random forest model, K-nearest neighbor model, and XGboost model.

[0018] Furthermore, the step of inputting the training samples into the classification model for iterative training until a preset number of iterations is reached to obtain an optimized classification model includes:

[0019] The training samples are divided into a training set and a test set;

[0020] The classification model is updated based on the training set, and the updated classification model is validated based on the test set.

[0021] Repeat the above steps until the optimized classification model is obtained.

[0022] Furthermore, this includes constructing the local BAM database, including:

[0023] Sequence alignment was performed between the reference genome and the corresponding raw sequencing data in the local gene database to obtain the BAM file;

[0024] The BAM files were sorted according to chromosome coordinates to remove duplicate sequences;

[0025] A local BAM database is constructed based on the deduplicated BAM files.

[0026] The system for assessing the influence of homologous regions in NGS gene variant detection includes:

[0027] The first acquisition module is used to acquire a number of first-class genes and a number of second-class genes based on a public database; wherein, the first-class genes are gene fragments whose mutation detection results have been confirmed to be affected by homologous regions, and the second-class genes are gene fragments whose mutation detection results have been confirmed to be unaffected by homologous regions.

[0028] The second acquisition module is used to acquire several original sequence files corresponding to the first type of gene and the second type of gene based on the local gene database, acquire sequence alignment files corresponding to each of the original sequence files based on the local BAM database, and combine each of the historical test data and the corresponding sequence alignment files to obtain several known NGS gene data.

[0029] The first construction module is used to calculate the GC content, the proportion of repetitive units, and the similarity ratio in the original sequence file, and to calculate the alignment quality in the corresponding sequence alignment file; and then to construct training samples corresponding to each of the known NGS gene data through these;

[0030] The training optimization module is used to input the training samples into the classification model for iterative training until the preset number of iterations is reached, with the goal of making the accuracy, precision, recall, F1 score and AUC value of the model evaluation parameters all approach 1, so as to obtain an optimized classification model.

[0031] The influence judgment module is used to extract GC content, repetitive unit ratio, similarity ratio and alignment quality from the NGS gene data to be detected and input them into the optimized classification model to determine whether the mutation detection of the corresponding gene is affected by the homologous region.

[0032] Furthermore, the training optimization module includes:

[0033] The training unit is used to simultaneously input the training samples into several different types of classification models to iteratively train them until a preset number of iterations is reached, so as to obtain several different types of intermediate classification models.

[0034] The selection unit is used to select the intermediate classification model whose AUC value is closest to 1 as the optimized classification model.

[0035] Furthermore, the training optimization module includes:

[0036] The preprocessing unit is used to divide the training samples into a training set and a test set;

[0037] An update unit is used to update the classification model based on the training set and to verify the updated classification model based on the test set.

[0038] The loop unit is used to repeatedly call the preprocessing unit and the update unit until the optimized classification model is obtained.

[0039] Furthermore, including:

[0040] The third acquisition module is used to perform sequence alignment between the reference genome and the corresponding raw sequencing data in the local gene database to obtain the BAM file;

[0041] The deduplication module is used to sort each of the BAM files according to chromosome coordinates to remove duplicate sequences;

[0042] The second construction module is used to build a local BAM database based on the deduplicated BAM file.

[0043] An electronic device, comprising:

[0044] At least one processor;

[0045] A memory that is communicatively connected to the at least one processor;

[0046] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the determination method.

[0047] Beneficial effects:

[0048] As can be seen from the above technical solutions, the technical solution of the present invention provides a method for judging the influence of homologous regions in NGS gene mutation detection, so as to improve the defects in the current judgment on whether homologous regions will affect NGS gene mutation detection.

[0049] The inventors discovered that, in order to determine whether homologous regions affect NGS gene mutation detection, current methods, based on traditional mathematical approaches and the inherent logic of the NGS platform itself (i.e., facilitating judgment through alignment), require sufficient prior information on the homologous regions of the target fragment. This prior information is then used to identify any detection anomalies caused by homologous regions during sequence alignment. However, not all homologous regions of target fragments possess sufficient prior information, and obtaining this prior information is complex, especially acquiring prior information on homologous regions of all target fragments in the genome, which requires significant costs. The inventors found that intelligent algorithms in the field of computer science can be trained with limited samples to automatically process an infinite amount of similar data. Therefore, computational learning was introduced into the process of determining the impact of homologous regions.

[0050] The method first obtains several first-class genes and several second-class genes whose influence on the detection of variants of homologous regions is known based on existing public databases, and then obtains several known NGS gene data corresponding to the first-class genes and the second-class genes.

[0051] In the process of constructing training samples, (1) First, considering the existence of GC preference in sequencing, regions with high / low GC content are less frequently measured, and the splicing of these regions is more difficult, which further affects a series of operations such as sequence alignment and variant detection. (2) Second, repetitive sequences play an indispensable role in gene expression and transcription regulation while influencing the evolution, heredity, and variation of life. However, the current sequencing technology produces relatively short reads. Furthermore, since the genome sequence is assembled using the kmer algorithm, highly similar repetitive sequences may be compressed together, affecting the identification of subsequent repetitive sequences. (3) Moreover, in biology, homology cannot be quantified. Similarity, on the other hand, can be used to describe the proportion of identical bases or amino acids between the detected sequence and the target sequence in the entire sequence. Generally, the higher the similarity between sequences, the greater the possibility of homology between sequences. (4) At the same time, a higher alignment quality score also indicates that the alignment to the current position is more reliable. In highly repetitive or homologous regions, shorter reads may not be unique when aligned with the reference genome, resulting in multiple mapping and read clustering. Furthermore, these regions also exhibit active recombination, which increases the difficulty of alignment and affects the alignment quality of these regions. Therefore, alignment quality serves as an important quantitative indicator reflecting the quality of sequence alignment to the reference genome, and is significantly affected by high repetition and homologous regions. It contains essential feature information for constructing a classification model that considers whether variant detection will be affected. Based on the above four considerations, we extracted GC content, repetitive unit ratio, similarity ratio, and alignment quality from known NGS gene data to construct training samples, ensuring the accuracy of the classification model training.

[0052] Multiple evaluation parameters were introduced as training objectives during the training of the classification model, further improving the reliability of the optimized model in classification judgments. Ultimately, with only the NGS gene data to be detected, it is possible to determine whether homologous regions will affect the mutation detection.

[0053] Therefore, this technical solution creatively introduces machine learning algorithms into gene sequencing and uses four gene parameters that have an important impact on the judgment results as training samples to achieve the training and optimization of the corresponding classification model. Thus, without needing to obtain sufficient prior information about the homologous region of the target fragment, it can accurately determine whether the homologous region will affect the gene variation detection under the NGS platform, thereby establishing a more accurate comparison and laying the foundation for subsequent phenotypic prediction and evolutionary analysis.

[0054] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below can be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other.

[0055] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0056] The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component shown in the various figures may be denoted by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the invention will now be described by way of example and with reference to the accompanying drawings, wherein:

[0057] Figure 1 This is a flowchart of the method for determining the influence of homologous regions in NGS gene variant detection as described in this embodiment;

[0058] Figure 2 A flowchart for obtaining type I and type II genes;

[0059] Figure 3 A flowchart for building a local BAM database;

[0060] Figure 4 Flowchart for optimizing the classification model;

[0061] Figure 5 A flowchart for selecting the optimal classification model. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art.

[0063] The terms "first," "second," and similar words used in the specification and claims of this patent application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, unless the context clearly indicates otherwise, the singular forms of "an," "a," or "the," etc., do not indicate a quantity limitation, but rather indicate the presence of at least one. Terms such as "comprising" or "including" mean that the element or object preceding "comprising" encompasses the features, wholes, steps, operations, elements, and / or components listed following "comprising" or "including," and do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or collections thereof. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0064] Researchers have reached a consensus that highly homologous sequences can significantly impact sequencing alignment and subsequent variant detection when performing gene variant detection using NGS platforms. However, without sufficient prior information about homologous regions within the target area, effective tools or methods are still lacking to assess questions such as: how high a sequence homology level is needed to affect variant detection, the extent of this impact, which gene fragments are affected by highly similar homologous regions, and which gene fragments are unaffected by homologous regions. Therefore, this embodiment aims to provide a method for determining the influence of homologous regions in NGS gene variant detection to improve upon the aforementioned shortcomings of existing technologies.

[0065] The following section, in conjunction with the accompanying drawings, details the method for determining the influence of homologous regions in NGS gene mutation detection provided in this embodiment.

[0066] Combination Figure 1 As shown, the method includes the following steps:

[0067] Step S102: Obtain several Class I genes and several Class II genes from a public database.

[0068] In this step, the first type of gene is a gene fragment whose mutation detection results have been confirmed to be affected by homologous regions, and the second type of gene is a gene fragment whose mutation detection results have been confirmed to be unaffected by homologous regions.

[0069] In practice, the publicly available databases are not limited to public gene databases and various literature sources. Specifically, the publicly available databases include NCBI, UCSC, Ensembl, etc.

[0070] like Figure 2 As shown, in one specific implementation, the first type of gene and the second type of gene are obtained through the following steps:

[0071] Step S102.2: Obtain gene location information that is affected by homologous regions and unaffected by homologous regions by using a coordinate extraction tool or a custom extraction program.

[0072] In this step, the location information is formatted as: chromosome number: gene start coordinates - gene end coordinates. For example, in the reference genome version GRCh38 of the human species, the coordinates of the BRCA1 gene are described as: chr17:43044294-43125364.

[0073] Step S102.4: Using a sequence extraction tool, gene fragments affected by homologous regions are obtained from the reference genome file based on the gene location information to form the first type of gene, and gene fragments not affected by homologous regions are formed into the second type of gene.

[0074] Step S104: Obtain several original sequence files corresponding to the first type of gene and the second type of gene based on the local gene database, obtain sequence alignment files corresponding to each of the original sequence files based on the local BAM database, and combine each of the historical test data and the corresponding sequence alignment files to obtain several known NGS gene data.

[0075] As a specific implementation method, such as Figure 3 As shown, the local BAM database mentioned in this step is created through the following steps:

[0076] Step S202: Perform sequence alignment between the reference genome and the corresponding raw sequencing data in the local gene database to obtain the BAM file.

[0077] The sequence alignment is specifically performed using a reference genome index file. Specifically, the reference genome index file is constructed through the following steps:

[0078] First, download the reference genome sequence in FASTA format (hereinafter referred to as ref.fa) from public gene databases (NCBI, UCSC, Ensembl, etc.);

[0079] Secondly, input the reference genome sequence file ref.fa, and use the sequence alignment software index building module to create an alignment index for the reference genome. The generated related files include ref.fa.fai, ref.fa.amb, ref.fa.ann, ref.fa.bwt, ref.fa.pac, and ref.fa.sa.

[0080] Step S204: Sort each of the BAM files according to chromosome coordinates to remove duplicate sequences.

[0081] Step S206: Construct a local BAM database based on the deduplicated BAM files.

[0082] Step S106: Calculate the GC content, repetitive unit ratio and similarity ratio in the original sequence file, and calculate the alignment quality in the corresponding sequence alignment file; then construct training samples corresponding to each of the known NGS gene data through them.

[0083] In this step, feature extraction from known NGS gene data to construct training samples is performed based on the following considerations:

[0084] Highly homologous regions, regions with high and low GC content, repetitive sequence regions, and poor alignment quality—these sequence limitations can all lead to insufficient coverage in NGS, thus affecting variant detection in NGS data. By collecting sequence characteristics that are highly correlated with or similar to homology and significantly affect variant detection, high-quality, high-utilization features can be encoded to provide strong support for obtaining more discriminative models. Specifically:

[0085] (1) Regarding GC content:

[0086] In fact, DNA molecules contain four bases: adenine (A), guanine (G), cytosine (C), and thymine (T). The ratio of guanine to cytosine among these four bases is called the GC content. Sequencing exhibits GC bias, meaning regions with a GC content of around 50% on the genome are more easily detected, generating more reads and offering greater coverage. Conversely, high-GC or low-GC regions are less easily detected, producing fewer reads and offering less coverage. Furthermore, due to this GC bias, high / low-GC regions are less frequently detected, making assembly more difficult and further impacting sequence alignment and variant detection.

[0087] In practice, the GC content is obtained by calculating the sequence base content at the gene position in the reference genome using bedtools software.

[0088] (2) Regarding the proportion of repeating units:

[0089] Repetitive sequences refer to identical or symmetrical segments appearing in the genome. "Identical" includes similar segments within the same genome or between genomes of different species. Numerous experiments have demonstrated that repetitive sequences contain a wealth of genetic information and are a crucial component of gene regulatory networks, playing an indispensable role in gene expression and transcriptional regulation while influencing evolution, heredity, and variation. Different types of repetitive sequences are important markers for species evolution research, playing a key role in chromosome structure and species evolution, contributing to human genetic variation, and exerting substantial impacts on human health through copy number variation, structural variation, insertion, deletion, and alterations in gene transcription and splicing. On the other hand, second-generation sequencing technologies currently produce relatively short reads. Because genome sequences are assembled using the k-mer algorithm, highly similar repetitive sequences may be compressed together, affecting the identification of subsequent repetitive sequences. Therefore, although repetitive regions play an important role in genome stability, due to technological limitations during the construction of the human reference genome GRCh38, these repetitive regions in the human reference genome were classified as blank or collapsed regions and thus ignored in many analytical experiments.

[0090] In practical implementation, since DNA sequences are categorized into single sequences and repetitive sequences based on their frequency of occurrence in the genome during feature encoding, RepeatMasker performs homology searches between the genome and a database of known repetitive sequences, masking any repetitive sequences identified in the genome as N or X. Therefore, based on the target sequence data FASTQ file, the RepeatMasker repetitive sequence detection tool accurately identifies or masks repetitive sequences in the genome by comparing them with a reference database, thus obtaining the characteristics of the proportion of repetitive regions in the sequence.

[0091] (3) Regarding the similarity ratio:

[0092] In biology, homology is used to describe evolutionary relationships between species. Therefore, homology expression can only use the terms "present" or "absent," making it a concept that cannot be quantified. Similarity, on the other hand, refers to the proportion of identical bases or amino acids between the detected sequence and the target sequence (a relatively macroscopic description). Sequence similarity allows comparison of the sequence under study with DNA or protein sequence libraries, which can then be used to determine the biological attributes associated with the sequence. Similar sequences often likely originate from a common ancestral sequence. They are likely to have similar spatial structures and biological functions. Generally, the higher the similarity between sequences, the greater the likelihood of homology. Therefore, using similarity to infer homology is feasible, and similarity can be considered a quantitative indicator of homology.

[0093] In practice, based on the target sequence data FASTQ file, overlapping k-mers in the query are searched in the genome to establish a list of matching hits between sequences. A method is designed to quickly discover sequences with a length of >= 40 bases and a similarity of >= 95%, and the similarity magnitude of the target sequence is obtained.

[0094] (4) Regarding the comparison quality:

[0095] The mapping quality (mapQ) of the alignment results was evaluated using a Phred scoring system similar to that used for base quality evaluation. For example, when mapQ = 20, it means that the error probability of aligning this read length to this position is 0.01. A higher mapping quality score indicates that the alignment to the current position is more reliable. Highly repetitive or homologous regions, and shorter reads, may not be unique when aligned with the reference genome, resulting in multiple mappings and read clustering. Furthermore, these regions also exhibit active recombination, which increases the alignment difficulty and affects the alignment quality of these regions. Therefore, alignment quality serves as an important quantitative indicator reflecting the quality of sequence alignment to the reference genome, and is significantly affected by highly repetitive and homologous regions. It contains essential feature information for constructing classification models that assess whether variant detection is affected.

[0096] In practice, the locally constructed BAM database contains a large number of alignment results of samples on the hg38 reference genome. By statistically analyzing the ratio of read counts with an alignment quality greater than 50 to the total number of reads in each sample file of the local BAM database for 308 target regions, the Mapping_quality index of the target sequence is obtained.

[0097] Therefore, this embodiment uses both of them to construct training samples.

[0098] Step S108: With the goal of making the accuracy, precision, recall, F1 score and AUC value in the model evaluation parameters all approach 1, the training samples are input into the classification model for iterative training until the preset number of iterations is reached, so as to obtain the optimized classification model.

[0099] like Figure 4 As shown, in one specific implementation method, the optimized classification model is obtained through the following steps:

[0100] Step S108.2: Divide the training samples into a training set and a test set.

[0101] Step S108.4: Update the classification model based on the training set, and verify the updated classification model based on the test set.

[0102] Step S108.6: Repeat the above steps until the optimized classification model is obtained.

[0103] like Figure 5 As shown, in order to improve the accuracy of subsequent judgments, this embodiment also optimizes the classification model. Based on this, step S108 specifically includes the following steps:

[0104] Step S108.2': Simultaneously input the training samples into several different types of classification models to iteratively train them until a preset number of iterations is reached, so as to obtain several different types of intermediate classification models.

[0105] In this step, the various types of classification models include: logistic regression model, support vector machine model, decision tree model, random forest model, K-nearest neighbor model, and XGboost model.

[0106] Step S108.4': Select the intermediate classification model with the AUC value closest to 1 as the optimized classification model.

[0107] Combining steps S108.2 to S108.6 and steps S108.2' to S108.4', in this embodiment, a total of 154 training samples corresponding to the first type of genes and 154 training samples corresponding to the second type of genes were obtained. The specific training results are shown in Table 1 below:

[0108] Table 1 shows the training results of each classification model.

[0109]

[0110] Generally, the AUC value is typically between 0.5 and 1.0. In this embodiment, the support vector machine model is ultimately chosen.

[0111] Step S110: Extract GC content, repetitive unit ratio, similarity ratio and alignment quality from the NGS gene data to be detected and input them into the optimized classification model to determine whether the mutation detection of the corresponding gene is affected by homologous regions.

[0112] Based on steps S102 to S110, the influence of homologous regions in variant detection can be determined without relying on prior information about homologous regions. Furthermore, the method described in this embodiment has low requirements for server computing resources; a typical 8-core, 64GB memory server can simultaneously run processing tasks for dozens of target genes. Simultaneously, this method has high versatility, low technical requirements for personnel, and can be accurately verified and judged even by personnel without any relevant technical background. Moreover, the process deployment of this method is simple and convenient to use; only the deployment of relevant computing nodes is needed to complete the entire analysis process.

[0113] The aforementioned program can run in a processor or be stored in memory (or a computer-readable storage medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media, such as modulated data signals and carrier waves.

[0114] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented using different modules, and different steps can be implemented using different modules.

[0115] This embodiment also provides a system for determining the influence of homologous regions in NGS gene variant detection. The system includes:

[0116] The first acquisition module is used to acquire a number of first-class genes and a number of second-class genes based on a public database; wherein, the first-class genes are gene fragments whose mutation detection results have been confirmed to be affected by homologous regions, and the second-class genes are gene fragments whose mutation detection results have been confirmed to be unaffected by homologous regions.

[0117] The second acquisition module is used to acquire several original sequence files corresponding to the first type of gene and the second type of gene based on a local gene database, acquire sequence alignment files corresponding to each of the original sequence files based on a local BAM database, and combine each of the historical test data and the corresponding sequence alignment files to obtain several known NGS gene data.

[0118] The first construction module is used to calculate the GC content, the proportion of repetitive units, and the similarity ratio in the original sequence file, and to calculate the alignment quality in the corresponding sequence alignment file; and then to construct training samples corresponding to each of the known NGS gene data through these.

[0119] The training optimization module is used to input the training samples into the classification model for iterative training until the preset number of iterations is reached, with the goal of making the accuracy, precision, recall, F1 score and AUC value of the model evaluation parameters all approach 1, so as to obtain an optimized classification model.

[0120] The influence judgment module is used to extract GC content, repetitive unit ratio, similarity ratio and alignment quality from the NGS gene data to be detected and input them into the optimized classification model to determine whether the mutation detection of the corresponding gene is affected by the homologous region.

[0121] The steps of the system used to implement the above method have already been explained and will not be repeated here.

[0122] For example, the training optimization module specifically includes:

[0123] The training unit is used to simultaneously input the training samples into several different types of classification models to iteratively train them until a preset number of iterations is reached, so as to obtain several different types of intermediate classification models.

[0124] The selection unit is used to select the intermediate classification model whose AUC value is closest to 1 as the optimized classification model.

[0125] For example, as an optional implementation, the training optimization module further includes:

[0126] The preprocessing unit is used to divide the training samples into a training set and a test set.

[0127] The update unit is used to update the classification model based on the training set and to verify the updated classification model based on the test set.

[0128] The loop unit is used to repeatedly call the preprocessing unit and the update unit until the optimized classification model is obtained.

[0129] For example, to enable the construction of a local BAM database, the system also includes:

[0130] The third acquisition module is used to perform sequence alignment between the reference genome and the corresponding raw sequencing data in the local gene database to obtain the BAM file.

[0131] The deduplication module is used to sort each of the BAM files according to chromosome coordinates to remove duplicate sequences.

[0132] The second construction module is used to build a local BAM database based on the deduplicated BAM file.

[0133] This embodiment also provides an electronic device. The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to perform the determination method described in this embodiment.

[0134] Since the judgment system and electronic equipment are both built on the judgment method, they can efficiently judge whether the homologous region will affect the mutation detection without prior information on the homologous region. In addition, the system and electronic equipment have the advantage of high throughput, and are also convenient to deploy, simple to implement and widely applicable.

[0135] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A method for judging the influence of homologous regions in NGS gene variant detection, characterized in that, The method comprises the following steps: Obtaining a plurality of first type genes and a plurality of second type genes according to a public database; wherein the first type genes are gene fragments whose variation detection results are confirmed to be affected by homologous regions, and the second type genes are gene fragments whose variation detection results are confirmed not to be affected by homologous regions; Obtaining a plurality of original sequence files corresponding to the first type genes and the second type genes based on a local gene database, obtaining a sequence alignment file corresponding to each original sequence file based on a local BAM database, and combining each original sequence file and the corresponding sequence alignment file to obtain a plurality of known NGS gene data; Calculating the GC content, repeat unit proportion and similarity ratio in the original sequence file, and calculating the alignment quality in the corresponding sequence alignment file; then constructing a training sample corresponding to each known NGS gene data through them; As the accuracy, precision, recall rate, F1 score and AUC value in the model evaluation parameter all tend to 1, inputting the training sample into a classification model for iterative training until a preset iteration number is reached to obtain an optimized classification model; Extracting the GC content, repeat unit proportion, similarity ratio and alignment quality from the NGS gene data to be detected and inputting them into the optimized classification model to determine whether the variation detection of the corresponding gene is affected by the homologous region.

2. The method of claim 1, wherein the homology region is determined by comparing the sequence of the target region of the test sample with the sequence of the target region of the reference sample. The method comprises the following steps: Inputting the training sample into a plurality of different types of classification models for iterative training until a preset iteration number is reached to obtain a plurality of different types of intermediate classification models; Selecting the intermediate classification model with the AUC value closest to 1 as the optimized classification model.

3. The method for judging the influence of homologous regions in NGS gene variation detection according to claim 2, wherein The plurality of different types of classification models comprise a logistic regression model, a support vector machine model, a decision tree model, a random forest model, a K-nearest neighbor model and an XGboost model.

4. The method for determining the influence of homologous regions in NGS gene variation detection according to claim 1, characterized in that, The method comprises the following steps: Dividing the training sample into a training set and a test set; Updating the classification model based on the training set and verifying the updated classification model based on the test set; Repeating the above steps until the optimized classification model is obtained.

5. The method of claim 1, wherein the homology region is determined by comparing the sequence of the read with the sequence of the reference genome. The method comprises the following steps of constructing the local BAM database: Performing sequence alignment between the reference genome and the corresponding original sequencing data in the local gene database to obtain a BAM file; Sorting each BAM file according to the chromosome coordinates to remove duplicate sequences; Constructing the local BAM database based on the de-duplicated BAM files.

6. A system for judging the influence of homologous regions in NGS gene variant detection, characterized by, The method comprises the following steps: The first obtaining module is configured to obtain a plurality of first genes and a plurality of second genes from a public database; the first genes are gene fragments whose variant detection results are affected by homologous regions, and the second genes are gene fragments whose variant detection results are not affected by homologous regions; The second obtaining module is configured to obtain a plurality of original sequence files corresponding to the first genes and the second genes based on a local gene database, obtain a sequence alignment file corresponding to each original sequence file based on a local BAM database, and combine each original sequence file and the corresponding sequence alignment file to obtain a plurality of known NGS gene data; The first constructing module is configured to calculate GC content, repeat unit proportion, and similarity ratio in the original sequence files, and calculate alignment quality in the corresponding sequence alignment files; and then construct a training sample corresponding to each known NGS gene data through the above parameters; The training optimization module is configured to input the training sample into a classification model for iterative training until a preset number of iterations is reached, so as to obtain an optimized classification model, with the accuracy, precision, recall rate, F1 score, and AUC value in the model evaluation parameters all tending to 1 as the target; The influence judgment module is configured to extract GC content, repeat unit proportion, similarity ratio, and alignment quality from the NGS gene data to be detected, and input them into the optimized classification model to determine whether the variant detection of the corresponding gene is affected by homologous regions. 7.The system for judging the impact of homologous regions in NGS genetic variation detection according to claim 6, wherein, The training optimization module comprises: A training unit configured to input the training sample into a plurality of different types of classification models for iterative training until a preset number of iterations is reached, so as to obtain a plurality of different types of intermediate classification models; A selection unit configured to select an intermediate classification model with an AUC value closest to 1 as the optimized classification model. 8.The system for judging the impact of homologous regions in NGS genetic variation detection according to claim 6, wherein, The training optimization module comprises: A preprocessing unit configured to divide the training sample into a training set and a test set; An updating unit configured to update a classification model based on the training set, and verify the updated classification model based on the test set; A loop unit configured to repeatedly call the preprocessing unit and the updating unit until the optimized classification model is obtained. 9.The system for judging the impact of homologous regions in NGS genetic variation detection according to claim 6, wherein, The method comprises: A third obtaining module configured to perform sequence alignment between a reference genome and corresponding original sequencing data in a local gene database to obtain a BAM file; A deduplication module configured to sort each BAM file according to chromosome coordinates to remove duplicate sequences; A second constructing module configured to construct a local BAM database based on the deduplicated BAM files.

10. An electronic device, comprising: The method comprises: At least one processor; A memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method for judging according to any one of claims 1 to 5.

Images