A model for high-frequency false positive regions and a method for filtering fixed-position mutations

Abstract

The present invention relates to the technical field of gene mutation detection, and specifically to a model of a high-frequency false positive region and a fixed position mutation filtering method, wherein the high-frequency false positive region model adopts a chi-square test statistic to obtain a mutation position distribution model based on a gene mutation result and takes the front 16bp and the back 20bp with significantly enriched mutations as the high-frequency false positive to-be-checked region, and the fixed position mutation filtering method obtains the gene mutation result based on contamination pre-filtering, false positive site filtering and true positive mutation judgment. At the same time, the method collects detection experimental data, continuously enriches the contamination pre-filtering database, and optimizes the true positive mutation judgment model, on the one hand, reduces the false positive incidence caused by the uneven distribution of the front and end mutations of reads, and on the other hand, continuously improves the accuracy of the true positive mutation judgment, and at the same time, more data support can be provided for subsequent research by saving the data.

Description

Technical Field

[0001] The present invention relates to the technical field of gene mutation detection, and in particular to a model of high-frequency false positive regions and a fixed position mutation filtering method. Background Art

[0002] Currently, the Next-Generation Sequencing (NGS) technology is becoming increasingly popular and has become an indispensable detection tool in the fields of genetics, biology, and clinical diagnosis. Among them, targeted sequencing technology is widely used in the study of disease-related genes and variant screening due to its high efficiency and cost-effectiveness. However, sequencing errors are often introduced during targeted sequencing, resulting in false positive mutations and affecting the accuracy of sequencing results.

[0003] False-positive mutations are usually caused by factors such as PCR amplification preferences, limitations of sequencing platforms, or contamination during sample processing. Among them, mutations at the front and end of reads are unevenly distributed. Such mutations are often highly abundant and cannot be filtered out by conventional filtering methods. In current applications, such false-positive results may lead to misdiagnosis or incorrect assessment of cancer, which has a profound negative impact on patient treatment and management. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a model of high-frequency false positive regions and a fixed position mutation filtering method, which can identify and eliminate high-frequency false positive regions at the start and end positions of reads, and ultimately achieve positive mutation judgment.

[0005] To achieve the above object, the present invention provides the following technical solution: a model of high-frequency false positive areas, the specific judgment logic of the model is as follows:

[0006] S101. Extract the gene sequencing data and compare it with the human reference genome hg19, and use the gene mutation detection tool to analyze and obtain the original mutation results;

[0007] S102. Count the original mutation results, and record the specific mutation position and frequency of each mutation on the reads;

[0008] S103. Using the chi-square test algorithm to calculate the significant distribution imbalance sites, record their positions on the reads, and construct a position distribution model;

[0009] S104. According to the position distribution model, the front 16bp and back 20bp where mutations are significantly enriched are selected as high-frequency false positive regions.

[0010] Furthermore, the present invention also provides a fixed position mutation filtering method, the method comprising:

[0011] S201. Using a sequencing tool to obtain raw gene sequencing data;

[0012] S202. Filter the sequence data of common laboratory contaminated microorganisms in gene mutation experiments by comparing the original gene sequencing data with the filtered data in the pre-filtered database;

[0013] S203. Using the above-mentioned model of a high-frequency false-positive region to calculate the high-frequency false-positive region to be checked;

[0014] S204. Based on the mutation positions and frequency data counted in S203, the mutation frequency distribution of the high-frequency false positive region to be investigated and the expected frequency distribution are evaluated to determine whether the site is a false positive site, as follows:

[0015] S20401. Count the positions and frequencies of mutations on reads and record them, where the total length of reads is recorded as L, the frequency of mutations in the front 16 bp of the high-frequency false positive region to be checked is recorded as A, the frequency of mutations in the back 20 bp of the high-frequency false positive region to be checked is recorded as C, and the frequency of mutations in the middle of the high-frequency false positive region to be checked is recorded as B;

[0016] S20402. Calculate the expected mutation distribution ratios of the front, middle, and back segments of reads, denoted as , , ;

[0017] S20403. Calculate the expected mutation frequencies at the beginning, middle, and end of reads, denoted as , , ,in is the total mutation frequency;

[0018] S20404. Apply the chi-square test to calculate the deviation between the observed mutation distribution and the expected mutation distribution to evaluate the mutation data distribution. The calculation formula is: ,when When it is greater than the threshold, the current site is determined to be a false positive site and the site is filtered;

[0019] S205. further performing true positive judgment on the remaining sites using the positive mutation judgment model to obtain the final gene mutation detection result;

[0020] S206. Collect the test data and test results and save them in the database;

[0021] S207. Compare the sequencing data after pre-filtration in S202 with the historical experimental data in the database, extract the sequence data that appears n times in the historical experimental data and put it into the pre-filter database as the common contaminating microbial sequence data of this experiment and this laboratory, where , m is the total number of historical experimental data, r is the set proportional factor;

[0022] S208. Generate the test report for this test.

[0023] In particular, the S202 pre-filtering database stores in advance the sequence data of common laboratory contamination microorganisms in gene mutation experiments.

[0024] In particular, the S205 positive mutation determination model is a supervised learning model trained using annotated data in a training database, wherein the annotated data in the training database is sequencing data with known mutation results in the database and valid determinations, and the corresponding mutation results serve as labels for the sequencing data.

[0025] Furthermore, the test data and test results saved in the database in S206 include key information of each main stage of the test experiment to support subsequent related analysis needs. The key information includes at least: the name of the main stage, the experimental control temperature, the step time, the experimental reagent type, the reagent dosage, the reagent concentration, and the reagent temperature.

[0026] Furthermore, step S207 also extracts the number of times N that the mutation data appears in the historical experimental data. When N is greater than the set threshold, step S208 also synchronously gives a key data recommendation report, which at least gives: repeated sequence data, key information of each main stage of this experiment, and key information of each main stage of the historical N experiments to prompt the experimenter to perform data analysis. The key information includes at least: the name of the main stage, the experimental control temperature, the step time, the type of experimental reagent, the reagent dosage, the reagent concentration, and the reagent temperature.

[0027] Furthermore, the database also stores reference values ​​of key information corresponding to each main stage, and the reference values ​​include at least: the experimental control temperature range, step operation time range, reagent type, reagent dosage range, reagent concentration range, and reagent use temperature range of this stage. When the deviation between the key information and the reference value in each main stage of the experiment is greater than the specified threshold, step S208 also simultaneously gives an experimental condition deviation report, and the report includes at least: the name of the main stage, the name of the deviation item, the reference value range of the deviation item, and the actual value of the deviation item.

[0028] In particular, each data saved in the database is annotated according to whether there are deviation items in the experimental process, and the data without deviation is judged as valid and added to the training database to iteratively train the positive mutation determination model.

[0029] Furthermore, the experimenters conducted data research and analysis based on the experimental report, key data recommendation report, and experimental condition deviation report, extracted valid data to iterate the positive mutation judgment model, improved the model recognition accuracy and generalization ability, extracted the data analyzed as contaminated and put it into the pre-filtering database to improve the S202 filtering quality.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] 1. The present invention optimizes the gene mutation detection process, reduces the incidence of false positives (especially false positives caused by uneven distribution of mutations at the front and end of reads), and significantly improves the credibility and clinical application value of sequencing data.

[0032] 2. The method of the present invention collects test results and continuously updates specific data. On the one hand, it updates the pre-filtering database by extracting public sequences. On the other hand, it continuously iterates and optimizes the gene mutation judgment model to improve the accuracy of the model gene mutation screening and the generalization ability of the model application. At the same time, the database collects intermediate information during the detection process, which can provide more data for subsequent research.

[0033] 3. For mutant gene sequences that appear N times repeatedly, when N > n, where n is the number of times the S207 sequence appears repeatedly in historical data, the sequence data is stored in the pre-filtering database, and a key data suggestion report is given to prompt the experimenter to further analyze the rationality of the data; when N ≤ n (but greater than the threshold), the sequence data is not stored in the pre-filtering database, but the data appears at a certain frequency, and a key data suggestion report is still given to remind the experimenter to pay attention to the data and clarify the meaning of the data to clarify whether the data is included in the pre-filtering database. Accordingly, the method of the present invention can further improve the accuracy of gene mutation detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A logical schematic diagram of the method of the present invention;

[0035] Figure 2 It is a logic diagram of the high-frequency false positive area model of the present invention;

[0036] Figure 3 This is a logic diagram of filtering false positive sites at fixed positions of the present invention;

[0037] Figure 4 This is a schematic diagram of the distribution of read mutations in the present invention;

[0038] Figure 5 This is a schematic diagram of the sequence of the high-frequency false positive region to be checked in the present invention;

[0039] Figure 6 This is a diagram for evaluating the experimental model by comparing the positive mutation judgment method alone with the method of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] Embodiment 1:

[0042] like Figure 1 As shown, after the method of the present invention obtains the test sample, it first uses a gene sequencer to perform gene sequencing (the second-generation sequencing is recommended for gene sequencing, which has high throughput, low cost, and sequencing accuracy> 99%) to obtain the offline data, and then compares the offline data with the data in the pre-filtering database. Since the pre-filtering database stores the sequence data of common laboratory contaminated microorganisms in gene mutation experiments in advance, the pre-filtering database also stores sequence data that appear repeatedly in multiple gene mutation detection processes (the repeated data may be the inherent product of a certain experimental operation, the product of a certain experimental condition, or the contaminated microorganism that is not completely treated in the laboratory) and the data obtained by the experimenters and researchers based on the experimental data. The data, experimental reports, key data suggestion reports, and experimental condition deviation reports are studied and analyzed, and the data that is ultimately identified as a source of contamination is finally identified. Therefore, through a pre-filtering operation, common contamination in this experiment can be filtered out, and the possibility of false positive results caused by sample contamination, etc. can be reduced. The filtered data is calculated by a high-frequency false positive region model to obtain a high-frequency false positive region to be checked, wherein this region especially includes a false positive region caused by uneven distribution of mutations at the front and end of reads. A false positive site judgment is performed on the high-frequency false positive region to be checked, wherein the false positive sites are filtered out, and the remaining site data are all input into the positive mutation judgment model for true positive mutation judgment to obtain the final gene mutation detection result. From the data processing level, after the original data is filtered to remove contamination, the false positive sites are removed by false positive filtering, and finally the true positive gene mutation results are further determined by the positive judgment model. Therefore, the gene mutation detection based on the present invention theoretically has a higher accuracy index; from the data experiment point of view, such as Figure 6As shown in the figure, it can be seen that the method of pre-filtering + false positive site filtering + positive mutation judgment of the present invention is significantly improved in AUC index (the larger the AUC value, the more likely the current classification algorithm is to rank positive samples ahead of negative samples, and the model has better classification performance) compared with the traditional judgment method based only on positive mutation model, indicating that the method of the present invention has obvious accuracy improvement. At the same time, the pre-filtering database and training database (database used for training true positive mutation judgment model) of the present invention are continuously updated as the experiment continues, so the filtering performance of the model of the present method and the performance of the mutation judgment model will be continuously optimized in the process of continuous iterative learning.

[0043] Embodiment 2:

[0044] like Figure 2 The high-frequency false positive area search model of the present invention is shown, and the specific steps are as follows:

[0045] 1. Compare the data from the next-generation sequencing machine with the human reference genome hg19, and then analyze the mutation results using a gene mutation detection tool (such as Strelka);

[0046] 2. Count the specific mutation positions and frequencies of each mutation on the reads, calculate the significant distribution imbalance sites through the chi-square test, record their positions on the reads, and construct a position distribution model, such as Figure 4 As shown;

[0047] 3. According to the position distribution model, the front 16bp and back 20bp where mutations are significantly enriched are used as high-frequency false positive areas to be checked.

[0048] Depend on Figure 4 It can be seen that the gene mutations at the start and end positions of reads are significant and high in frequency, which are usually false positive mutations caused by PCR amplification preferences and sequencing platform limitations. In application, they should be removed to prevent misdiagnosis.

[0049] Embodiment three:

[0050] like Figure 3 , Figure 5 As shown, the present invention performs false positive site screening on a fixed area of ​​a high-frequency false positive area, wherein the site judgment is based on the assumption that the mutation is evenly distributed over the entire reads, and the specific steps are as follows:

[0051] 1. Count the mutation positions and frequencies in high-frequency false positive regions. In this case, the total length of reads is L=143:

[0052] a. The frequency of the first 16 bp is A = 96;

[0053] b. Frequency of middle 107bp: B = 30;

[0054] c. Frequency of the back 20 bp: C = 121;

[0055] 2. Calculate the expected mutation distribution ratios in the front, middle, and back segments of reads, , , ;

[0056] 3. Calculate the expected mutation frequency in the front, middle and back segments of the reads, where the total mutation frequency T = A + B + C = 247:

[0057] a. Expected mutation in the first 16 bp: ;

[0058] b. Expected mutation in the middle 107bp: ;

[0059] c. Expected mutation in the back 20bp: ;

[0060] 4. Calculate the chi-square value: , take the significance level of 0.05, the degree of freedom is 2, Table 1 is the critical value table of chi-square distribution, from Table 1 below, we can see that the significant critical value is 5.99, >5.99, so the site in this case is a false positive site.

[0061] Table 1 Chi-square distribution critical value table

[0062]

[0063] Embodiment 4:

[0064] The present invention collects experimental data of each mutation detection experiment, and the data include experimental conclusions, experimental process information, and experimental parameter comparison information, wherein the experimental process information includes reagents, reagent dosage, reagent concentration, reagent temperature, experimental control temperature, and experimental operation time corresponding to each process; the experimental parameter comparison information is the difference between the experimental process information and the theoretical range of the parameters. If there is a difference, the difference information needs to be recorded. At the same time, at the end of the experiment, an experimental condition deviation report is generated while generating an experimental report to inform the experimenter that there is a problem with the experimental condition control, and whether the experimental conclusion is correct can be further evaluated.

[0065] When multiple detection experiments are completed, the collected data will be compared with the historical data. If some sequence data is repeated, the number of repetitions is recorded as N. When N>n, where n is the number of times the S207 sequence appears repeatedly in the historical data, the sequence data is stored in the pre-filtering database, and a key data suggestion report is given to prompt the experimenter to further analyze the rationality of the data; when N≤n (but greater than the threshold), the sequence data is not stored in the pre-filtering database, but the data appears at a certain frequency, and a key data suggestion report is still given to remind the experimenter to pay attention to the data and clarify the meaning of the data to clarify whether the data is included in the pre-filtering database. Accordingly, the method of the present invention can further improve the accuracy of gene mutation detection.

[0066] All data are stored in the database, and when researchers need to study, they can call it out randomly. Since the data fully records the intermediate process parameters and conclusions of each experimental data, researchers can directly use the data for analysis, discover the data rules, and further explore more knowledge about gene sequences based on the data to promote the development of biomedicine. At the same time, in the method of the present invention, further research on the data rules found, such as finding more pollution source data and more data belonging to mutation conclusions, can be used to further improve the filtering effect of the pre-filtering model and the predictive performance of the positive judgment model.

[0067] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fixed position mutation filtering method, Features: Methods include: S201. Using a sequencing tool to obtain raw gene sequencing data; S202. Filter the sequence data of common laboratory contaminated microorganisms in gene mutation experiments by comparing the original gene sequencing data with the filtered data in the pre-filtered database; S203. The high-frequency false positive region to be checked is obtained by using the model calculation of the high-frequency false positive region. The specific judgment logic of the model is as follows: S101. Extract the gene sequencing data and compare it with the human reference genome hg19, and use the gene mutation detection tool to analyze and obtain the original mutation results; S102. Count the original mutation results, and record the specific mutation position and frequency of each mutation on the reads; S103. Using the chi-square test algorithm to calculate the significant distribution imbalance sites, record their positions on the reads, and construct a position distribution model; S104. According to the position distribution model, the front 16bp and the back 20bp where the mutation is significantly enriched are selected as the high-frequency false positive region; S204. Based on the mutation positions and frequency data counted in S203, the mutation frequency distribution of the high-frequency false positive region to be investigated and the expected frequency distribution are evaluated to determine whether the site is a false positive site, as follows: S20401. Count the positions and frequencies of mutations on reads and record them, where the total length of reads is recorded as L, the frequency of mutations in the front 16 bp of the high-frequency false positive region to be checked is recorded as A, the frequency of mutations in the back 20 bp of the high-frequency false positive region to be checked is recorded as C, and the frequency of mutations in the middle of the high-frequency false positive region to be checked is recorded as B; S20402. Calculate the expected mutation distribution ratios of the front, middle, and back segments of reads, denoted as , , ; S20403. Calculate the expected mutation frequencies at the beginning, middle, and end of reads, denoted as , , ,in is the total mutation frequency; S20404. Apply the chi-square test to calculate the deviation between the observed mutation distribution and the expected mutation distribution to evaluate the mutation data distribution. The calculation formula is: ,when When it is greater than the threshold, the current site is determined to be a false positive site and the site is filtered; S205. further performing true positive judgment on the remaining sites using the positive mutation judgment model to obtain the final gene mutation detection result; S206. Collect the test data and test results and save them in the database; S207. Compare the sequencing data after pre-filtration in S202 with the historical experimental data in the database, extract the sequence data that appears n times in the historical experimental data and put it into the pre-filter database as the common contaminating microbial sequence data of this experiment and this laboratory, where , m is the total number of historical experimental data, r is the set proportional factor; S208. Generate the test report for this test.

2. A fixed position mutation filtering method as claimed in claim 1, Features: The S202 pre-filtering database stores in advance the sequence data of common laboratory contamination microorganisms in gene mutation experiments.

3. A fixed position mutation filtering method as claimed in claim 2, Features: The S205 positive mutation determination model is a supervised learning model trained using the annotated data in the training database, wherein the annotated data in the training database is sequencing data with known mutation results in the database and valid determinations, and the corresponding mutation results serve as labels for the sequencing data.

4. A fixed position mutation filtering method as claimed in claim 3, Features: The test data and test results saved in the database in S206 include key information of each main stage in the test experiment process to support subsequent related analysis needs. The key information includes at least: the name of the main stage, the experimental control temperature, the step time, the experimental reagent type, the reagent dosage, the reagent concentration, and the reagent temperature.

5. A fixed position mutation filtering method as claimed in claim 4, Features: In step S207, the number of times N that the mutation data appears in the historical experimental data is extracted. When N is greater than the set threshold, a key data suggestion report is also given synchronously in step S208. The report at least gives: repeated sequence data, key information of each main stage of this experiment, and key information of each main stage of the historical N experiments to prompt the test experimenter to perform data analysis. The key information includes at least: main stage name, experimental control temperature, step time, experimental reagent type, reagent dosage, reagent concentration, and reagent temperature.

6. A fixed position mutation filtering method as claimed in claim 5, Features: The database also stores reference values ​​of key information corresponding to each main stage, and the reference values ​​include at least: the experimental control temperature range, step operation time range, reagent type, reagent dosage range, reagent concentration range, and reagent use temperature range of this stage. When the deviation between the key information and the reference value in each main stage of the experiment is greater than the specified threshold, step S208 also simultaneously gives an experimental condition deviation report, and the report includes at least: the name of the main stage, the name of the deviation item, the reference value range of the deviation item, and the actual value of the deviation item.

7. A fixed position mutation filtering method as claimed in claim 6, Features: Each time the data is saved in the database, it is annotated according to whether there are deviation items in the experimental process. The data without deviation is judged as valid and added to the training database to iteratively train the positive mutation determination model.

8. A fixed position mutation filtering method as claimed in claim 7, Features: The experimenter conducts data research and analysis based on the experimental report, key data recommendation report, and experimental condition deviation report, extracts valid data to iterate the positive mutation judgment model, improves the model's recognition accuracy and generalization ability, extracts data analyzed as contaminated and puts it into the pre-filtering database to improve the S202 filtering quality.

Images