CN115101133A - Integrated learning-based SNP interaction detection system - Google Patents

Abstract

The utility model provides a SNP interaction detecting system based on ensemble learning, which belongs to the technical field of artificial intelligence data mining classification and bioinformatics, and the scheme comprises: a data acquisition module configured to: acquiring SNP sequence information of a diseased sample and a non-diseased sample, and carrying out pretreatment to realize the construction of a whole genome SNP set; a SNP subset partitioning and combination generation module configured to: dividing a genome-wide SNP set into a plurality of SNP subsets, and constructing SNP combinations based on the SNP subsets; a multi-classifier parallel evaluation module configured to: evaluating the association of the SNP combination with the disease in parallel using a plurality of classifiers; a result verification module configured to: statistical significance validation was performed using the chi-square test on SNP combinations associated with disease assessed using several classifiers.

Description

A SNP Interaction Detection System Based on Ensemble Learning

technical field

The present disclosure belongs to the field of artificial intelligence data mining classification and bioinformatics technology, and in particular relates to an integrated learning-based SNP interaction detection system.

Background technique

The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

With the development of whole-genome sequencing and high-throughput technologies, researchers have been able to obtain information on hundreds of millions of single nucleotide polymorphisms (SNPs). However, how to use appropriate machine learning technology to find the combination of multiple SNPs related to disease from massive SNP data is a difficulty that still needs to be solved in the application of current machine learning technology in association detection.

The existing methods for SNP combinations associated with diseases include: using statistical methods such as hypothesis testing to study the significance of the association between each SNP combination and disease; using SNP information to divide samples into multiple subsets, according to the results of dividing apples and diseases The correlation of the whole genome SNP data is firstly clustered by using the clustering algorithm, and then the related SNP combinations are found within each cluster. This method can significantly reduce the candidate set and reduce the computational burden.

The inventor found that due to the high dimensionality of SNP data, and the number of combinations increases exponentially with the increase of the dimensionality, there are problems such as excessive detection burden, difficult to achieve, and prone to high false positive rate, which makes the detection of multiple SNPs difficult. There is still much room for improvement in machine learning techniques for the study of associations between interactions and diseases or traits.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present disclosure provides a SNP interaction detection system based on ensemble learning. The scheme divides the SNP set into multiple subsets, selects SNP combinations that may be related to the disease in the subsets, and further iteratively Selecting more relevant SNP combinations reduces the required memory space and running time; the system uses multiple classifiers to evaluate together, which can reduce the impact of different classifiers' preferences on disease models on the overall effect of the algorithm; using multiple classifiers The parallel detection of the classifiers improves the detection speed and reduces the hardware requirements of the system.

According to a first aspect of the embodiments of the present disclosure, a system for detecting SNP interactions based on ensemble learning is provided, including:

a data acquisition module, which is configured to: acquire the SNP sequence information of the diseased samples and the non-diseased samples, and perform preprocessing to realize the construction of a genome-wide SNP set;

A SNP subset division and combination generation module is configured to: divide a genome-wide SNP set into a plurality of SNP subsets, and construct a SNP combination based on the SNP subsets;

A multi-classifier parallel evaluation module, which is configured to: use multiple classifiers to evaluate the correlation between the SNP combination and the disease in parallel;

A result validation module configured to: use a chi-square test to validate the statistical significance of combinations of disease-related SNPs evaluated using several classifiers.

Further, the whole genome SNP set is divided into multiple SNP subsets, and a high-dimensional SNP combination is constructed based on the SNP subset, taking the two-point SNP combination as an example, specifically:

Divide the genome-wide SNP collection evenly into multiple SNP subsets;

In the first iteration process, for each SNP subset, two different SNPs are selected to form a two-site SNP combination, and all possible SNP combinations in the subset form a set;

In the second and subsequent iterations, for every two different SNP subsets, one SNP is selected in each subset to form a two-site SNP combination; all possible SNP combinations between the two subsets constitute a set; And input the possible disease-related SNP combination output in the previous iteration process into the SNP combination set that has not been detected, as the input of the classifier in this iteration process.

Further, the multi-classifier parallel evaluation module includes a scoring voting module, an exchange voting module and a screening module, wherein:

A scoring voting module, which is configured to: use each classifier to score the input SNP combination, and vote according to the score;

an exchange voting module, which is configured to: exchange the SNP combinations that each classifier thinks may be related to the disease to all other classifiers, and repeat the scoring voting;

The screening module is configured to: count the voting situation of each classifier, screen out the SNP combination whose total number of votes is greater than the preset threshold, and input it to the result verification module.

Further, the chi-square test is used to verify the statistical significance of disease-related SNP combinations obtained by using several classifiers, specifically:

p-values were calculated for all combinations of SNPs considered to be disease-related by the multiclassifier using the chi-square test;

Sort these SNP combinations in ascending order according to their p-values;

Find a p-value inflection point and output the SNP combination before the inflection point as the final detection result.

Further, the SNP sequence information of the diseased sample and the non-diseased sample is obtained, and preprocessed to realize the construction of the whole genome SNP set, specifically:

The diseased sample is marked as 1; the unaffected sample is marked as 0; for the mutation of the SNP data sample at each SNP site, if both alleles are not mutated, it is marked as 0; If one of the alleles is mutated, it is marked as 1; if both alleles are mutated, it is marked as 2; if the data at this locus is missing, it is marked as 3; at the same time, delete the number of missing SNPs Samples greater than 5%; SNPs with more than 5% missing samples were deleted; p-values for each SNP were calculated using the chi-square test, and SNPs with p-values > 0.0001 were deleted; SNPs with less than 0.1 minor allele frequencies were deleted.

Further, the plurality of classifiers include classifiers based on gini index, k2-score, entropy, information gain and APDS.

According to a second aspect of the embodiments of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the memory, where the processor implements the following steps when executing the program:

Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections;

dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets;

Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated;

Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

According to a third aspect of the embodiments of the present disclosure, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the following steps are implemented:

Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections;

dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets;

Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated;

Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

According to a fourth aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program that, when run on one or more processors, performs the following steps:

Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections;

dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets;

Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated;

Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

Compared with the prior art, the beneficial effects of the present disclosure are:

(1) The present disclosure provides a SNP interaction detection system based on ensemble learning. The scheme divides the SNP set into multiple subsets, selects SNP combinations that may be related to the disease in the subsets, and further iteratively selects more The combination of related SNPs reduces the required memory space and running time; the system uses multiple classifiers to evaluate together, which can reduce the influence of the preferences of different classifiers on the disease model on the overall effect of the algorithm; the use of multiple classifiers Parallel detection improves the detection speed and reduces the hardware requirements of the system.

(2) The scheme described in this disclosure incorporates all possible SNP combinations into the scope of association evaluation, avoids omitting SNP combinations that are significantly related to the disease, and enhances the reliability of the algorithm results; at the same time, multiple classifiers are used to combine SNPs The correlation of the classifier is evaluated, which reduces the influence of a single classifier on the model preference on the overall result of the algorithm, and multiple classifiers can be executed in parallel on multiple devices, reducing the computational burden and the requirements for the experimental environment.

(3) The scheme described in this disclosure divides the whole genome SNP set into multiple SNP subsets, and gradually evaluates SNP combinations in multiple iterations, instead of directly evaluating all possible SNP combinations at one time, which reduces the storage space of the device. Requirements; according to the inflection point of the p-value of the chi-square test, the SNP set that is significantly related to the disease is divided, rather than a rigid division boundary, which reduces the influence of parameter settings on the experimental results.

Advantages of additional aspects of the disclosure will be set forth in part in the description that follows, and in part will become apparent from the description below, or will be learned by practice of the disclosure.

Description of drawings

The accompanying drawings that constitute a part of the present disclosure are used to provide further understanding of the present disclosure, and the exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure.

FIG. 1 is a schematic diagram of the overall flow of the SNP interaction detection system based on ensemble learning described in the embodiment of the present disclosure.

Detailed ways

The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the present disclosure. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that when the terms "comprising" and/or "including" are used in this specification, it indicates that There are features, steps, operations, devices, components and/or combinations thereof.

The embodiments of this disclosure and features of the embodiments may be combined with each other without conflict.

Example 1:

The purpose of this embodiment is to provide a SNP interaction detection system based on ensemble learning.

An ensemble learning-based SNP interaction detection system, comprising:

a data acquisition module, which is configured to: acquire the SNP sequence information of the diseased samples and the non-diseased samples, and perform preprocessing to realize the construction of a genome-wide SNP set;

A SNP subset division and combination generation module is configured to: divide a genome-wide SNP set into a plurality of SNP subsets, and construct a SNP combination based on the SNP subsets;

A multi-classifier parallel evaluation module, which is configured to: use multiple classifiers to evaluate the correlation between the SNP combination and the disease in parallel;

A result validation module configured to: use a chi-square test to validate the statistical significance of combinations of disease-related SNPs evaluated using several classifiers.

Further, the SNP subset division and combination generation module specifically includes a SNP subset division module and a combination generation module; wherein:

SNP subset partitioning module, which is configured as:

Divide the genome-wide SNP collection into multiple SNP subsets;

The composition builds the module, which is configured as:

Construct a high-dimensional SNP set based on a subset of SNPs; fuse the SNP combinations considered to be disease-related in the previous iterative process and the SNP combinations that have not been evaluated as the input of this iterative process.

Further, the multi-classifier parallel evaluation module includes a scoring voting module, an exchange voting module and a screening module; wherein:

A scoring and voting module, which is configured to: use a single classifier to score SNP combinations, and vote according to the scores;

an exchange voting module, which is configured to: exchange the SNP combinations that a single classifier thinks may be related to the disease to all other classifiers, and repeat the scoring voting;

The screening model is configured to: count the voting situation of each classifier, filter out the SNP combination with a high total number of votes, and input it to the combination generation module.

Further, obtain SNP sequence information of diseased samples and non-diseased samples, and perform data preprocessing;

Among them, the diseased sample is marked as 1; the unaffected sample is marked as 0;

Among them, the mutation of the sample at each SNP site in the SNP data, if neither alleles are mutated, it is marked as 0; if one of the two alleles is mutated, it is marked as 1 ; If both alleles are mutated, mark 2; if data at this locus is missing, mark 3;

Among them, the specific steps of data preprocessing are: delete samples with missing SNPs greater than 5%; delete SNPs with missing samples greater than 5%; use chi-square test to calculate the p value of each SNP, delete SNPs with p value>0.0001; delete SNPs with minor allele frequencies less than 0.1.

Further, the whole genome SNP set is divided into multiple SNP subsets and SNP combinations are generated, taking the two-point SNP combination as an example, specifically including:

S1021, evenly dividing the whole genome SNP set into multiple SNP subsets;

S1022, in the first iteration process, for each SNP subset, select two different SNPs to form a two-site SNP combination, and all possible SNP combinations in the subset form a set;

S1023, in the second and subsequent iterations, for every two different SNP subsets, select one SNP in each subset to form a two-site SNP combination; all possible SNP combinations between the two subsets constitute a and input the possible disease-related SNP combination output from the previous iteration process into a SNP combination set that has not been detected yet, as the input of a classifier in this iteration process.

Further, use multiple classifiers to evaluate multiple SNP combination sets in parallel; specifically including:

为第t次迭代过程中，输入到第i个分类器的SNP组合，i∈{1,…,|D|}，拟采用|D|＝5个分类器进行检测：gini指数，k2-score，熵，信息增益，APDS(AbsoluteProbability Different Score)；具体的：make

is the SNP combination input to the i-th classifier during the t-th iteration, i∈{1,...,|D|}, and |D|=5 classifiers are to be used for detection: gini index, k2-score , entropy, information gain, APDS (AbsoluteProbability Different Score); specific:

其中N是所有样本的数目，N_case(m)为具有基因型m且表现为患病个体的数目，N_control(m)为具有基因型m且表现为不患病个体的数目，N_total(m)为具有基因型m的样本的数目。分数越高代表SNP组合与疾病相关性越强；S1031, the Gini index (Gini index, GI) calculates the score for each SNP combination as GI=Gini(parent)-Gini(split), and

where N is the number of all samples, N _case (m) is the number of individuals with genotype m and manifested as having the disease, N _control (m) is the number of individuals with genotype m and manifested as not having the disease, N _total ( m) is the number of samples with genotype m. The higher the score, the stronger the correlation between the SNP combination and the disease;

为了简便计算，我们取其log形式：

分数越高代表SNP组合与疾病相关性越强；S1032, K2-score calculates the score for each SNP combination as

For simplicity of calculation, we take its log form:

The higher the score, the stronger the correlation between the SNP combination and the disease;

其中

分数越高代表SNP组合与疾病相关性越强；S1033, Entropy score (ES) calculates the score for each SNP combination as

in

The higher the score, the stronger the correlation between the SNP combination and the disease;

分数越高代表SNP组合与疾病相关性越强；S1034, the information gain (Information Gain, IG) calculates the score for each SNP combination as: IG=[H(S _i |Y)+H(S _j |Y)-H(S _i ,S _j |Y)]- [H(S _i )+H(S _j )-H(S _i ,S _j )], and

The higher the score, the stronger the correlation between the SNP combination and the disease;

分数越高代表SNP组合与疾病相关性越强；S1035, APDS calculates the score for each SNP combination as

The higher the score, the stronger the correlation between the SNP combination and the disease;

的SNP组合被认为可能与疾病相关，并对这些SNP组合在当前分类器下的投票数更新为

其中b_u是预先设定的参数，o是排序后的顺序，

是本次迭代过程输入到该分类器中的SNP组合个数。由于各个分类器相互独立，输入组合也互不影响，因此该过程可以并行执行；S1036, after each classifier completes the scoring of all SNP combinations input into it, the SNP combinations are sorted in descending order according to the scores, and the sorted order is less than

of SNP combinations considered likely to be disease-related, and the number of votes for these SNP combinations under the current classifier is updated as

where b _u is a preset parameter, o is the sorted order,

is the number of SNP combinations input into the classifier in this iteration process. Since the classifiers are independent of each other and the input combinations do not affect each other, the process can be executed in parallel;

其中，j∈D/i，且j₁∪j₂∪…∪j_|D|-1∪i＝D。对新组合集合的打分投票方式同上一步相同；S1037, exchange the SNP combinations considered to be possibly related to the disease in each classifier to all other classifiers, and the set of SNP combinations to be evaluated after the exchange is:

where j∈D/i, and j ₁ ∪j ₂ ∪…∪j _|D|-1 ∪i=D. The scoring and voting method for the new combination set is the same as the previous step;

若仍有SNP组合集合未被任何一个分类器检测过，则将

输入至组合生成模块，进入下一次迭代过程；否则，将

输出至结果验证模块；S1038 , count the votes of each SNP combination by each classifier, and use the combination whose total number of votes is greater than V _S as the output of this iteration process

If there is still a SNP combination set that has not been detected by any classifier, the

Input to the combination generation module to enter the next iteration process; otherwise, the

Output to the result verification module;

Further, the chi-square test was used to further verify the evaluation results of multiple classifiers, and the SNP combination with a significant correlation with the disease was selected; specifically:

S1041, use chi-square test to calculate its p-value for all SNP combinations considered to be disease-related by the multi-classifier;

S1042, sort these SNP combinations in ascending order according to the p value;

S1043, find a p-value inflection point, and consider the SNP combination before the inflection point to be significantly related to the disease, as the final output result of the algorithm.

The present disclosure divides the genome-wide SNP set into multiple SNP subsets, then generates multiple high-dimensional SNP combination sets from these subsets, and then inputs several SNP combination sets into a multi-classifier, and performs parallel analysis on each SNP in the set Scores are calculated in combination, and votes are made according to the scores, and the SNP combination with high score under each classifier is exchanged to all other classifiers for scoring and voting again, and then the SNP combination with a high total number of votes from multiple classifiers is considered to be likely to be related to the disease. Relevant, input into the next iterative process or enter the next step, use the chi-square test to verify the degree of statistical correlation between these SNP combinations and the disease, and finally output the SNP combinations that are significantly related to the disease as the final result of the algorithm. The present disclosure avoids missing truly relevant combinations by dividing the SNP set into multiple subsets and repeating multiple iterations until all SNP combinations have been detected. In each iteration process, the detection process of multiple classifiers can be performed in parallel, reducing The reduced computational burden makes it possible to perform an exhaustive enumeration of all possible cases in the whole genome. In addition, the algorithm uses multiple classifiers to jointly determine whether a certain SNP combination is related to disease, which alleviates the influence of a single classifier's preference on the data model on the algorithm results. Finally, the chi-square test is used for further verification, which increases the reliability of the algorithm results. reliability, and did not use hard indicators to classify whether it was related to disease, which reduced the impact of parameter settings on the results.

Embodiment 2:

The purpose of this embodiment is to provide an electronic device.

An electronic device, comprising a memory, a processor and a computer program stored on the memory to run, the processor implements the following steps when executing the program:

Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections;

dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets;

Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated;

Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

Further, the steps executed by the electronic device described in this embodiment are consistent with the solution executed by the system described in Embodiment 1, and the technical details thereof have been described in detail in Embodiment 1, so they are not repeated here.

Embodiment three:

The purpose of this embodiment is to provide a non-transitory computer-readable storage medium.

A non-transitory computer-readable storage medium on which a computer program is stored, the program implements the following steps when executed by a processor:

Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections;

dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets;

Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated;

Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

Further, the steps executed by the non-transitory computer-readable storage medium described in this embodiment are consistent with the solution executed by the system described in Embodiment 1, and the technical details thereof have been described in detail in Embodiment 1, so they are not described here. Repeat.

Embodiment 4:

The purpose of this embodiment is to provide a computer program product.

A computer program product comprising a computer program that, when run on one or more processors, performs the following steps:

Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections;

dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets;

Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated;

Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

Further, the steps performed by the computer program product described in this embodiment are consistent with the solution performed by the system described in Embodiment 1, and the technical details thereof have been described in detail in Embodiment 1, so they are not repeated here.

The SNP interaction detection system based on ensemble learning provided by the above embodiments can be implemented and has broad application prospects.

The above descriptions are only preferred embodiments of the present disclosure, and are not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included within the protection scope of the present disclosure.

Claims (10)

1. a SNP interaction detection system based on ensemble learning, is characterized in that, comprises: a data acquisition module, which is configured to: acquire the SNP sequence information of the diseased samples and the non-diseased samples, and perform preprocessing to realize the construction of a genome-wide SNP set; A SNP subset division and combination generation module is configured to: divide a genome-wide SNP set into a plurality of SNP subsets, and construct a SNP combination based on the SNP subsets; A multi-classifier parallel evaluation module, which is configured to: use multiple classifiers to evaluate the correlation between the SNP combination and the disease in parallel; A result validation module configured to: use a chi-square test to validate the statistical significance of combinations of disease-related SNPs evaluated using several classifiers. 2. a kind of SNP interaction detection system based on ensemble learning as claimed in claim 1, is characterized in that, described dividing whole genome SNP collection into multiple SNP subsets, and based on described SNP subsets construct high-dimensional The SNP combination, taking the two-point SNP combination as an example, is as follows: Divide the genome-wide SNP collection evenly into multiple SNP subsets; In the first iteration process, for each SNP subset, two different SNPs are selected to form a two-site SNP combination, and all possible SNP combinations in the subset form a set; In the second and subsequent iterations, for every two different SNP subsets, one SNP is selected in each subset to form a two-site SNP combination; all possible SNP combinations between the two subsets constitute a set; And input the possible disease-related SNP combination output in the previous iteration process into the SNP combination set that has not been detected, as the input of the classifier in this iteration process. 3. a kind of SNP interaction detection system based on ensemble learning as claimed in claim 1, is characterized in that, described multi-classifier parallel evaluation module comprises scoring voting module, exchange voting module and screening module, wherein: A scoring voting module, which is configured to: use each classifier to score the input SNP combination, and vote according to the score; an exchange voting module, which is configured to: exchange the SNP combinations that each classifier thinks may be related to the disease to all other classifiers, and repeat the scoring voting; The screening module is configured to: count the voting situation of each classifier, screen out the SNP combination whose total number of votes is greater than the preset threshold, and input it to the result verification module. 4. a kind of SNP interaction detection system based on ensemble learning as claimed in claim 1, it is characterized in that, described using chi-square test to adopt some classifiers to evaluate and obtain and carry out statistical analysis on the SNP combination that obtains related to disease. Significance verification, specifically: p-values were calculated for all combinations of SNPs considered to be disease-related by the multiclassifier using the chi-square test; Sort these SNP combinations in ascending order according to their p-values; Find a p-value inflection point and output the SNP combination before the inflection point as the final detection result. 5. a kind of SNP interaction detection system based on ensemble learning as claimed in claim 1, is characterized in that, described obtaining the SNP sequence information of diseased sample and non-diseased sample, and carry out preprocessing, realize whole genome SNP The construction of the collection, specifically: The diseased sample is marked as 1; the unaffected sample is marked as 0; for the mutation of the SNP data sample at each SNP site, if both alleles are not mutated, it is marked as 0; If one of the alleles is mutated, it is marked as 1; if both alleles are mutated, it is marked as 2; if data is missing for this locus, it is marked as 3. 6. The system for detecting SNP interaction based on ensemble learning according to claim 5, wherein the preprocessing further comprises: deleting samples with more than 5% missing SNPs; deleting samples with more than 5% missing samples SNPs; p-values for each SNP were calculated using the chi-square test, SNPs with p-values > 0.0001 were deleted; SNPs with minor allele frequencies less than 0.1 were deleted. 7 . The SNP interaction detection system based on ensemble learning according to claim 1 , wherein the multiple classifiers comprise classifiers based on gini index, k2-score, entropy, information gain and APDS. 8 . 8. An electronic device, comprising a memory, a processor and a computer program stored on the memory, wherein the processor implements the following steps when executing the program: Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections; dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets; Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is zero, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated. Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point. 9. A non-transitory computer-readable storage medium on which a computer program is stored, characterized in that, when the program is executed by a processor, the following steps are implemented: Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections; dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets; Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated; Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point. 10. A computer program product, comprising a computer program, characterized in that the computer program, when run on one or more processors, performs the following steps: Obtain SNP data of diseased individuals and non-diseased individuals, and perform preprocessing to realize the construction of genome-wide SNP collections; dividing the genome-wide SNP set into multiple SNP subsets, and constructing SNP combinations based on the SNP subsets; Input some SNP combinations into the classifier for voting; at the same time, exchange the SNP combination with strong disease correlation under each classifier to all other classifiers, and evaluate the vote again; filter out the total number of votes higher than the preset threshold The number of SNP combinations obtained is cleared, mixed with the SNP combinations that have not been evaluated by the classifier, and the above steps are repeated until all SNP combinations have been evaluated; Based on the chi-square test, the SNP combination with the total number of votes screened out in the last iteration process is higher than the preset threshold, find the data inflection point from the p-value sequence, and output the SNP combination before the inflection point.

Images