AU2023261122A1

AU2023261122A1 - Construction method for model for analyzing variation detection result

Info

Publication number: AU2023261122A1
Application number: AU2023261122A
Authority: AU
Inventors: Zhiyu Peng; Jun Sun; Fei Tang; Zhonghua Wang
Original assignee: Bgi Tianjin; Tianjin Medical Laboratory Bgi
Current assignee: Bgi Tianjin; Tianjin Medical Laboratory Bgi
Priority date: 2022-04-25
Filing date: 2023-03-15
Publication date: 2024-09-05
Also published as: WO2023207396A1; CN116994647A

Abstract

The present invention provides a construction method for a model for analyzing a variation detection result. The method comprises: obtaining a positive sequencing data set specified for a positive variation site and a negative sequencing data set specified for a negative variation location; respectively extracting features of the variation sites from the positive sequencing data set and the negative sequencing data set; and constructing a model by using the feature result obtained in the previous step, wherein the features comprise at least one of the following: an AD0 value, an AD1 value, an AF0 value, an AF1 value, a GT value, a DP value, a GQ value, an MQ value, and a QUAL value.

Description

CONSTRUCTION METHOD FOR MODEL FOR ANALYZING VARIATION DETECTION RESULT FIELD

[0001] The present application relates to the field of biology. In particular, the present

application relates to a method for constructing a model for analyzing variation detection result.

BACKGROUND

[0002] Clinical next generation sequencing (cNGS) is widely used to determine the

molecular diagnosis of patients with genetic diseases. However, known NGS procedures suffer

from random and systematic errors in the steps of sequencing, alignment, and variation

invoking. As the reported variations can affect patient care and treatment, the American College

of Medical Genetics and Genomics (ACMG) and the College of American Pathologists (CAP)

recommend orthogonal validation of the reported variations to reduce the risk of false positive

results. Currently, Sanger sequencing is the main technique for molecular diagnosis of genetic

diseases. However, as evidenced by the growth of public databases such as ClinVar and OMIM,

the total number of candidate variations for clinical reporting is steadily increasing, which

multiplies the cost and turnover time of testing. Thus, it is impractical to completely detect the

candidate variations. Therefore, it is becoming increasingly urgent to reduce the need for

orthogonal testing by identifying false positive variations in cNGS data based on machine

learning models, which are trained with a large number of known data.

[0003] At present, there are the following problems in the study regarding variation false

positive: orthogonal experiments such as Sanger sequencing may substantially increase the

costs and turnover time; most of the features used in the existing models are Boolean tag values,

which may lead to information loss compared with the unchanged quantitative indicators; there

are relatively few false positive variations invoking in the existing model training set, which may lead to a wider confidence interval of some false positive capture rates (especially SNV); existing models use insufficient clinical data due to the cost, either deliberately complex for multiple scenarios with insufficient confidence, or sufficient confidence with a high risk of overfitting, inadequate for scenarios.

[0004] Therefore, it is urgent to improve the current methods for predicting variation false positives .

SUMMARY

[0005] The present disclosure aims to solve, at least to some extent, at least one of the technical problems existing in the prior art.

[0006] In an aspect, the present disclosure provides a method for constructing a model for analyzing variation detection result. According to an embodiment of the present disclosure, the

method includes: acquiring a positive sequencing data set of identified positive variation sites

and a negative sequencing data set of identified negative variation sites; extracting features of variation sites from the positive sequencing data set and the negative sequencing data set; and

constructing a model based on features and results acquired in the previous step. The features include at least one of: ADO value representing a depth of the first allele in the variation site

genotype; AD1 value representing a depth of the second allele in the variation site genotype;

AFO value representing a frequency of the first allele in the variation site genotype; AF1 value of representing a frequency of the second allele in the variation site genotype; GT value

representing a single numerical value (specifically, 0, 1, 2, or 3); DP value representing a

sequencing depth value; GQ value representing a quality value of variation site genotype; MQ value representing a quality of mapping of variation sites; and QUAL value representing a

quality value of probability of the variation sites.

[0007] Several dozens of result parameters can be generated by the variation detection and analysis software. The Applicant compared and analyzed these result parameters, and screened

out a group of result parameters. By using these result parameters as features a machine learning model is constructed for the data sets of the identified positive and negative variation sites. With

the obtained model, it can be accurately predicted whether the positive variation data are false positive, and the variation site genotype can be learned, thereby quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal experiment.

[0008] In another aspect, the present disclosure provides a method for analyzing variation detection result. According to an embodiment of the present disclosure, the method includes:

predicting, by analyzing the candidate positive variation data set with a machine learning model obtained using the above-mentioned method for constructing the model for analyzing variation detection result, whether positive variation data in the candidate positive variation data set are

false positive and/or variation site genotype. In this way, using the method according to the

present disclosure, it can be accurately predicted whether the positive variation data are false positives using the methods of the present disclosure, while the variation genotype can be

determined, thereby quickly and precisely locating the possible variations and reducing the cost

and turnover time of orthogonal experiment.

[0009] In yet another aspect, the present disclosure provides an apparatus for constructing a model for analyzing variation detection result. According to an embodiment of the present

disclosure, the apparatus includes: an acquisition module configured to acquire a positive sequencing data set of identified positive variation sites and a negative sequencing data set of

identified negative variation sites; an extraction module configured to extract features of

variation sites from the positive sequencing data set and the negative sequencing data set; and a construction module configured to construct a model based on features and results acquired

by the extraction module. The features include at least one of: ADO value representing a depth of the first allele in the variation site genotype; ADI value representing a depth of the second allele in the variation site genotype; AFO value representing a frequency of the first allele in the

variation site genotype; AF Ivalue of representing a frequency of the second allele in the variation site genotype; GT value representing a single numerical value; DP value representing

a sequencing depth value; GQ value representing a quality value of variation site genotype; MQ

value representing a quality of mapping of variation sites; and QUAL value representing a quality value of probability of the variation sites. Thus, by using the model obtained by the

apparatus according to the present disclosure, it can be accurately predicted whether the positive

variation data are false positive, while the variation genotype can be determined, thereby quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal experiment.

[0010] In yet another aspect, the present disclosure provides an executable storage medium. According to an embodiment of the present disclosure, the storage medium has computer

program instructions stored thereon. The computer program instructions, when run on a

processor, cause a processor to execute the method for analyzing variation detection result as described above. Thus, by executing the storage medium according to the present disclosure, it can be accurately predicted whether the positive variation data are false positive, while the

variation genotype can be determined, thereby quickly and precisely locating the possible

variations and reducing the cost and turnover time of orthogonal experiment.

[0011] In yet another aspect, the present disclosure provides an electronic device.

According to an embodiment of the present disclosure, the electronic device includes the above

mentioned executable storage medium and a processor configured to execute a computer program to implement the above-mentioned method for analyzing variation detection result.

Thus, by implementing the electronic device of the present disclosure, it can be accurately

predicted whether the positive variation data are false positive, while the variation genotype can be determined, thereby quickly and precisely locating the possible variations and reducing

the cost and turnover time of orthogonal experiment.

[0012] Additional aspects and advantages of the present disclosure will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be

learned by practice of the present disclosure.

DETAILED DESCRIPTION

[0013] Hereinafter, embodiments of the present disclosure are described in detail. The embodiments described below are illustrative only and are not to be construed as limitations of the present disclosure. The specific techniques or conditions, when not specified in the

examples, are performed according to techniques or conditions described in literatures in the

related art or according to the product instructions. The used reagents or instruments without indicating the manufacturer are conventional products that are commercially available.

Method for constructing a model for analyzing variation detection result

[0014] In an aspect, the present disclosure provides a method for constructing a model for analyzing variation detection result. According to an embodiment of the present disclosure, the method includes: acquiring a positive sequencing data set of identified positive variation sites

and a negative sequencing data set of identified negative variation sites; extracting features of

variation sites from the positive sequencing data set and the negative sequencing data set; and constructing a model based on features and results acquired in the previous step. The features include at least one of ADO value, ADi value, AFO value, AFi value, GT value, DP value, GQ

value, MQ value, and QUAL value.

[0015] Through a large number of experiments, the Applicant screened out nine result parameters as mentioned above, which are all the result parameters in the GATK software. The

specific meanings thereof can refer to the following table. These result parameters are used as

features to perform machine learning on the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation

sites to obtain a prediction model. Thus, by using the obtained model, it can be accurately

predicted whether the positive variation data are false positives, thereby quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal

experiment.

[0016] [Table 1] Meanings of features Features Meanings

site genotype (e.g., if ADO Allele depth of the first allele in the variation GT = 0/1, it represents a depth of reference allele) site genotype AD1 Allele depth of the second allele in the variation (e.g., if GT = 0/1, it represents a depth of the first candidate allele) site genotype AF0 Allele frequency of the first allele in the variation (e.g., GT = 0/1 and AD = (10, 30), then AFO value is 0.25)

variation site genotype AF1 Allele frequency of the second allele in the (e.g., GT = 0/1 and AD = (10,30), then AFIvalue is 0.75)

value, GT Genotype field (GT) is converted to a single numerical where GT value can be 0, 1, 2, or 3 (e.g., GT = 0/1, GT value is 0) DP Sequencing depth value GQ Quality value of variation site genotype

MQ Quality of mapping of variation sites QUAL Quality value of the probability of variation sites

[0017] According to an embodiment of the present disclosure, the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified

negative variation sites are acquired by: acquiring a sequencing data set; aligning, by using GATK software, the sequencing data set with reference data to acquire a candidate positive

variation data set; and performing analysis processing on the candidate positive variation data

set to acquire the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites.

[0018] Preferably, the clinical gene sequencing data are first acquired, and the sequencing data are compared with the reference data (including operations such as alignment, variation detection, annotation, and filtration); the variations are identified with GATK to acquire

candidate positive variation data; and VCF file is output. By performing the analysis processing on the candidate positive variation data again, it can be clearly determined whether the data are

true positive or false positive. The data are divided into the positive sequencing data set for the

positive variation sites and the negative sequencing data set for the negative variation sites.

[0019] According to an embodiment of the present disclosure, the reference data is selected

from human genome hg19.

[0020] According to an embodiment of the present disclosure, the analysis processing includes: performing standard clinical interpretation on the candidate positive variation data set

to acquire a data set of potentially pathogenic variations; and performing orthogonal test

analysis on the data set of potentially pathogenic variations to acquire the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the

identified negative variation sites. The positive sequencing data set includes an SNV variation

type data set and anTNDEL variation type data set. Each of the SNV variation type data set and the INDEL variation type data set includes a homozygous genotype data set and a heterozygous

genotype data set.

[0021] The term "standard clinical interpretation" refers to an interpretation of the pathogenicity of clinical variants with reference to the ACMG guideline, edition 2015.

[0022] Based on standard clinical interpretation of the candidate positive variation data acquired by GATK identification analysis, the potentially pathogenic variation data can be acquired, and then the accuracy of variation can be verified through orthogonal test on these data, thereby acquiring the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites. The positive sequencing data set can be divided into SNV variation type and INDEL variation type. The genotype of variation, i.e., Hom or Het, can be accurately known based on these two variation types.

[0023] It should be noted that the method for orthogonal test analysis according to the present disclosure is not strictly limited as long as it can be known whether the variation data of possible pathogenic diseases is true positive variation or false positive. It can be specifically

manipulated by using conventional techniques in the art, for example, refer to Sanger F. DNA

sequencing with chain-terminating inhibitors, 1977[J]. Biotechnology (Reading, Mass.), 24:104-108.

[0024] According to an embodiment of the present disclosure, the model is selected from a random forest classification model and has a threshold of 0.95 0.05. The setting of the threshold ensures a sufficient accuracy rate and reduces accidental errors. With the scalable

threshold setting, the accuracy rate and the orthogonal test rate can be balanced on the premise

of ensuring sufficient accuracy.

[0025] According to an embodiment of the present disclosure, the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites are divided into a training set and a test set (3:1), respectively, and a random forest classification model is selected, and the model with the highest accuracy rate is

selected through 5-fold cross validation.

[0026] According to an embodiment of the present disclosure, the method for constructing a model for analyzing variation detection result includes the following steps.

[0027] 1. First, clinical genomic data are acquired and aligned with the human reference genome (hg19), and GATK is used to identify the variation and output VCF file.

[0028] 2. The potentially pathogenic variations are acquired based on standard clinical interpretation, then the accuracy of variation through the orthogonal experiment is verified, and the accurate genotypes of homozygous (Hom), heterozygous (Het), and no variation (N) are provided.

[0029] 3. The VCF file is then converted to machine learning labels, from which a total of 9 features were acquired, see Table 1.

[0030] 4. Depending on the different variation types (SNV, INDEL), two different machine learning classification models are constructed by the features extracted from the VCF file and the optimal parameters are acquired by grid search.

[0031] 5. The data are respectively divided into a training set and a test set (3:1) based on the above method, and a random forest classification model is selected, and the model with the

highest accuracy rate is selected through 5-fold cross validation. Method for analyzing variation detection result

[0032] In another aspect, the present disclosure provides a method for analyzing variation detection result. According to an embodiment of the present disclosure, the method includes: acquiring a candidate positive variation data set; and predicting, by analyzing the candidate

positive variation data set with a machine learning model, whether positive variation data in the

candidate positive variation data set are false positive and/or variation site genotype, the machine learning model being acquired by the above-mentioned method for constructing the

model for analyzing variation detection result. In this way, by using the model obtained by the

method according to the present disclosure, it can be accurately predicted whether the candidate positive variation data are false positives, while the variation genotype can be determined,

thereby quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal experiment.

[0033] According to an embodiment of the present disclosure, the candidate positive variation data set is acquired by: acquiring a sequencing data set; and aligning, by using GATK software, the sequencing data set with reference data to acquire the candidate positive variation

data set.

[0034] According to an embodiment of the present disclosure, the model is selected from a random forest classification model, and a confidence of the candidate positive variation data is

lower than a threshold of the model, the candidate positive variation data is subjected to an

orthogonal test analysis, to predict whether the positive variation data in the candidate positive variation data set are false positive. The data lower than the threshold are referred to as gray zone data, and the accuracy rate of predicting false positives using the model is low. Therefore, this part of the data is necessarily subjected to orthogonal experimental verification to accurately predict whether it is false positive.

[0035] Those skilled in the art can appreciate that the features and advantages described above with respect to the method for constructing a model for analyzing variation detection result are equally applicable to the method for analyzing variation detection result, which are not be described in detail herein.

Apparatus for constructing a model for analyzing variation detection result

[0036] In yet another aspect, the present disclosure provides an apparatus for constructing a model for analyzing variation detection result. According to an embodiment of the present

disclosure, the apparatus includes: an acquisition module configured to acquire a positive

sequencing data set of identified positive variation sites and a negative sequencing data set of identified negative variation sites; an extraction module configured to extract features of

variation sites from the positive sequencing data set and the negative sequencing data set; and

a construction module configured to construct a model based on features and results acquired by the extraction module. The features includes at least one of: ADO value representing a depth

of the first allele in the variation site genotype; ADI value representing a depth of the second

allele in the variation site genotype; AFO value representing a frequency of the first allele in the variation site genotype; AF Ivalue of representing a frequency of the second allele in the

variation site genotype; GT value representing a single numerical value; DP value representing a sequencing depth value; GQ value representing a quality value of variation site genotype; MQ value representing a quality of mapping of variation sites; and QUAL value representing a

quality value of probability of the variation sites. Thus, with the model obtained by the apparatus according to the present disclosure, it can be accurately predicted whether positive

variation data are false positives, while the variation genotype can be determined, thereby

quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal experiment.

[0037] According to an embodiment of the present disclosure, the acquisition module includes: a sequencing data set acquisition module configured to acquire a sequencing data set; an aligning module configured to align, by using GATK software, the sequencing data set with reference data to acquire a candidate positive variation data set; and an analysis processing module configured to perform analysis processing on the candidate positive variation data set to acquire the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites. The acquisition module can accurately determine the positive variation site data and negative variation site data in the sequencing data set, and can also determine the genotype of the positive variation site.

[0038] According to an embodiment of the present disclosure, the analysis processing module includes a standard clinical interpretation module configured to perform standard

clinical interpretation on the positive variation data to acquire data of potentially pathogenic variations; and an orthogonal test analysis module configured to perform orthogonal test

analysis on the data set of potentially pathogenic variations to acquire the positive sequencing

data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites.

Executable storage medium

[0039] In yet another aspect, the present disclosure provides an executable storage medium. According to an embodiment of the present disclosure, the storage medium has computer

program instructions stored thereon. The computer program instructions, when executed on a

processor, cause the processor to implement the method for analyzing variation detection result as described above. Thus, with the storage medium of the present disclosure, it can be accurately

predicted whether positive variation data are false positives, while the variation genotype can be determined, thereby quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal experiment.

[0040] Those skilled in the art can appreciate that the features and advantages described above in relation to the method for analyzing variation detection result are equally applicable

to the executable storage medium, which are not be described further herein.

Electronic device

[0041] In yet another aspect, the present disclosure provides an electronic device. According to an embodiment of the present disclosure, the electronic device includes the

executable storage medium described above, and a processor configured to execute the computer program to implement the method for analyzing variation detection result described above. Thus, by implementing the electronic device of the present disclosure, it can be accurately predicted whether positive variation data are false positives, while the variation genotype can be determined, thereby quickly and precisely locating the possible variations and reducing the cost and turnover time of orthogonal experiment.

[0042] Those skilled in the art can appreciate that the features and advantages described above with respect to the method for analyzing variation detection result and the executable storage medium are equally applicable to the electronic device, which are not be described in

detail herein.

[0043] The embodiments of the present disclosure will be explained with reference to the following examples. It will be understood by those skilled in the art that the following examples

are merely illustrative of the present disclosure and are not to be construed as limiting the scope

of the present disclosure. Where specific techniques or conditions are not specified in the examples, they are performed according to techniques or conditions described in the literature

in the art or according to the product description. The reagents or instruments used without

indicating the manufacturer are conventional and commercially available products. Example 1

[0044] 1. WES data of 5190 clinical patients were acquired, and GATK software was used to perform alignment, variation detection, annotation, and filtration on the data and human genome hg19 to acquire a VCF file.

[0045] 2. The VCF file was subjected to the standard clinical interpretation process, to acquire 7375 potentially pathogenic variants by analysis.

[0046] 3. The above-mentioned 7375 variants numbers were verified by orthogonal experiments (for details, referring to Sanger F. DNA sequencing with chain-terminating inhibitors. , 1977[J]. Biotechnology (Reading, Mass.), 24:104-108), and it was determined that

these variations included 5241 variation type SNV and 2134 variation type INDEL. There were

3226 Het genotypes, 63 Hom genotypes, and 1952 negative variants in SNV; and there were 1606 Het genotypes, 138 Hom genotypes, and 390 negative variants in Indel.

[0047] 4. The data in the previous step was divided into a training set and a test set (3:1), the training set established a random forest classification model respectively, and all the features in the training set were taken as candidate features. Then, the principal component analysis was performed, and finally, 9 features listed in Table 2 were determined.

[0048] [Table 2] Feature importance in establishing random forest classification model with different variation types SNV and INDEL

Features SNV_MODEL INDELMODEL

ADO 0.0305 0.0389

ADI 0.0365 0.0606

AFO 0.3350 0.3135

AFI 0.2352 0.3027

GT 0.0078 0.0289

DP 0.0300 0.0174

GQ 0.0787 0.0691

MQ 0.0139 0.0141

QUAL 0.2324 0.1548

[0049] The test set accuracy rate of SNV and INDEL models was 94.8% and 93.8%, respectively, and the accuracy rates of different genotypes are shown in Table 3.

[0050] [Table 3] Accuracy rate of different genotypes in establishing random forest classification model with different variation types SNV and INDEL

Genotype SNVMODEL(%) INDELMODEL(%)

Het 92.9 80.5

Horn 100 92.1

N (negative) 96.3 97.2

[0051] Considering the required accuracy of clinical data, this method acquires different accuracy and orthogonal experiment ratio (see Table 4) by defining different thresholds

(confidence of random forest results) for the test set. The accuracy rate represents the number

of correct judgment/the total number meeting the threshold, and the orthogonal experiment ratio represents the number lower than the threshold/the number of overall test samples. Under the

condition that sufficient accuracy rate is satisfied, the smallest possible threshold of orthogonal experiment ratio is selected as the target threshold, and the threshold is finally determined to be

0.95, within a scalable range 0.05. The results indicate that the proposed method has certain

tolerance to noise data, data redundancy, and low-quality data, with excellent robustness.

[0052] [Table 4] Different thresholds and ratio of required orthogonal experiments in establishing random forest classification models with different variation types SNV and INDEL

Orthogonal experiment Variation type Threshold Accuracy rate(%) ratio(%)

Default (the maximum of Het, hom, SNV 94.8 0 and N (negative))

SNV 0.9 98.1 14.8 SNV 0.92 98.3 17.2

SNV 0.94 98.6 20.8

SNV 0.96 98.7 26.7 SNV 0.99 99.1 87.1 Default (the maximum of Het, hom, INDEL 93.8 0 and N (negative))

INDEL 0.9 97.8 21.1

INDEL 0.92 98.0 23.7

INDEL 0.94 98.0 27.8 INDEL 0.96 97.7 34.4

INDEL 0.99 98.1 69.5

[0053] In the specification, references to descriptions of the terms "an embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. mean that a particular feature, structure, material, or characteristic described in connection with the

embodiment or example is included in at least one embodiment or example of the present

disclosure. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features,

structures, materials, or characteristics described may be combined in any suitable manner in

any one or more embodiments or examples. Furthermore, combinations and combinations of the various embodiments or examples and features of the various embodiments or examples

described in this specification can be made by those skilled in the art without departing from

the scope of the present disclosure.

[0054] While the embodiments of the present disclosure are illustrated and described above, it will be understood that the above-described embodiments are illustrative and not restrictive and that those skilled in the art can make changes, modifications, substitutions, and alterations without departing from the scope of the present disclosure.

Claims

What is claimed is: 1. A method for constructing a model for analyzing variation detection result, the method

comprising:

acquiring a positive sequencing data set of identified positive variation sites and a negative sequencing data set of identified negative variation sites;

extracting features of variation sites from the positive sequencing data set and the negative sequencing data set; and

constructing a model based on features and results acquired in the previous step,

wherein the features comprise at least one of: ADO value: a depth of a first allele in a variation site genotype;

ADi value: a depth of a second allele in the variation site genotype;

AFO value: a frequency of the first allele in the variation site genotype; AFi value: a frequency of the second allele in the variation site genotype;

GT value: a single numerical value; DP value: a sequencing depth value;

GQ value: a quality value of the variation site genotype;

MQ value: a quality of mapping of the variation sites; and QUAL value: a quality value of probability of the variation sites.
2. The method according to claim 1, wherein the positive sequencing data set of the

identified positive variation sites and the negative sequencing data set of the identified negative variation sites are acquired by:

acquiring a sequencing data set; aligning, by using GATK software, the sequencing data set with reference data to acquire

a candidate positive variation data set; and

performing analysis processing on the candidate positive variation data set to acquire the positive sequencing data set of the identified positive variation sites and the negative

sequencing data set of the identified negative variation sites.
3. The method according to claim 2, wherein the reference data is selected from human

genome hg19.
4. The method according to claim 2, wherein the analysis processing comprises: performing standard clinical interpretation on the candidate positive variation data set to

acquire a data set of potentially pathogenic variations; and performing orthogonal test analysis on the data set of potentially pathogenic variations to acquire the positive sequencing data set of the identified positive variation sites and the negative

sequencing data set of the identified negative variation sites, wherein the positive sequencing

data set comprises an SNV variation type data set and an INDEL variation type data set, each of the SNV variation type data set and the INDEL variation type data set comprising a

homozygous genotype data set and a heterozygous genotype data set.
5. The method according to claim 1, wherein the model is selected from a random forest classification model and has a threshold of 0.95 0.05.
6. A method for analyzing variation detection result, comprising:

acquiring a candidate positive variation data set; and predicting, by analyzing the candidate positive variation data set with a machine learning

model, whether positive variation data in the candidate positive variation data set are false

positive and/or variation site genotype, the machine learning model being acquired by the method for constructing the model for analyzing variation detection result according to any one

of claims I to 5.
7. The method according to claim 6, wherein the candidate positive variation data set is acquired by:

acquiring a sequencing data set; and aligning, by using GATK software, the sequencing data set with reference data to acquire

the candidate positive variation data set.
8. The method according to claim 6, wherein: the model is selected from a random forest classification model; and

when a confidence of the candidate positive variation data is lower than a threshold of the

model, the candidate positive variation data is subjected to an orthogonal test analysis, to predict whether the positive variation data in the candidate positive variation data set are false positive.
9. A apparatus for constructing a model for analyzing variation detection result, comprising:

an acquisition module configured to acquire a positive sequencing data set of identified positive variation sites and a negative sequencing data set of identified negative variation sites;

an extraction module configured to extract features of variation sites from the positive

sequencing data set and the negative sequencing data set; and a construction module configured to construct a model based on features and results acquired by the extraction module,

wherein the features comprise at least one of:

ADO value: a depth of a first allele in a variation site genotype; ADi value: a depth of a second allele in the variation site genotype;

AFO value: a frequency of the first allele in the variation site genotype;

AFi value: a frequency of the second allele in the variation site genotype; GT value: a single numerical value;

DP value: a sequencing depth value;

GQ value: a quality value of the variation site genotype; MQ value: a quality of mapping of the variation sites; and

QUAL value: a quality value of probability of the variation sites.
10. The apparatus according to claim 9, wherein the acquisition module comprises: a sequencing data set acquisition module configured to acquire a sequencing data set;

an aligning module configured to, by using GATK software, the sequencing data set with reference data to acquire a candidate positive variation data set; and an analysis processing module configured to perform analysis processing on the candidate

positive variation data set to acquire the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites.
11. The apparatus according to claim 10, wherein the analysis processing module

comprises: a standard clinical interpretation module configured to perform standard clinical

interpretation on the positive variation data to acquire a data set of potentially pathogenic

variations; and an orthogonal test analysis module configured to perform orthogonal test analysis on the data set of potentially pathogenic variations to acquire the positive sequencing data set of the identified positive variation sites and the negative sequencing data set of the identified negative variation sites.
12. An executable storage medium, having computer program instructions stored thereon,

wherein the computer program instructions, when executed on a processor, cause the processor to implement the method for analyzing variation detection result according to any one of claims 6 to 8.
13. An electronic device, comprising:

the executable storage medium according to claim 12; a processor configured to execute a computer program to implement the method for

analyzing variation detection result according to any one of claims 6 to 8.