Genetic risk prediction model construction method integrating functional annotation information and application

Abstract

The invention belongs to the technical field of genetic risk prediction, and discloses a functional annotation information integrated genetic risk prediction model construction method and application, and the method comprises the steps: calculating the genetic degree of the specific annotation of each cell type corresponding to each tissue of a to-be-detected sample based on the marginal chi-square statistic # imgabs0 # of each SNP; the joint effect size bf of the M SNPs in the tissue f is estimated according to the genetic degree # imgabs1 # of the jth SNP in the current tissue f, the optimal linear unbiased estimation # imgabs2 # of the bf is obtained based on the covariance between the bf and phenotype vectors y of the n to-be-detected samples, and then a genetic risk prediction model # imgabs3 # is obtained; and incorporating the function annotation information of each organization into genetic risk prediction. The invention further provides a genetic risk prediction method integrating the function annotation information. The accuracy of genetic risk prediction can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of genetic risk prediction, and more specifically, relates to a method for constructing a genetic risk prediction model integrating functional annotation information and its application. Background Art

[0002] Genetic risk score (PRS), as an important method for disease risk prediction in systems epidemiology, can be used for population risk prediction, optimization of screening programs and precise prevention.

[0003] PRS involves screening SNPs (single nucleotide polymorphisms) and evaluating the true effects of the screened SNPs. In the prior art, some PRS methods further screen and estimate effects by integrating genetically related traits and functional annotation information. For example, by directly incorporating SNP functional annotations into the prior distribution of effect size, SNPs are screened and the true effects of SNPs are estimated based on the prior distribution. However, this method is directly based on the heritability of all SNPs, resulting in low accuracy in the final genetic risk prediction. Summary of the invention

[0004] In response to the above defects or improvement needs of the prior art, the present invention provides a method for constructing a genetic risk prediction model integrating functional annotation information and its application, the purpose of which is to improve the accuracy of genetic risk prediction.

[0005] To achieve the above object, according to a first aspect of the present invention, a method for constructing a genetic risk prediction model is provided, comprising:

[0006] S1, marginal chi-square statistic based on the jth SNP Calculate the heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested, and then obtain the heritability of the jth SNP in the current tissue f Where j is 1, 2, ... M, and M represents the total number of SNPs in each tissue of the sample to be tested; Indicates the cell type corresponding to tissue f The heritability of specific annotation, c takes 1, 2, ... C, C represents the total number of cell types corresponding to tissue f;

[0007] S2, using the random effects model y = Xb f +ε f Estimate the joint effect size b of M SNPs in tissue f f ; where y is the n-dimensional phenotype vector of the sample to be tested, X is the M*n-dimensional genotype matrix, ε f is the n-dimensional residual vector of organization f, n is the total number of samples to be tested;

[0008] S3, based on bf The covariance between y and b f The optimal linear unbiased estimate of Among them, D f and Respectively represent b f and ε f The covariance matrix of σ 2 Pick u is a constant;

[0009] S4. Constructing a genetic risk prediction model PRS f It represents the genetic risk score corresponding to tissue f, and the level of the genetic risk score is used to reflect the level of genetic risk.

[0010] Furthermore, after obtaining the genetic risk scores corresponding to each tissue, it also includes:

[0011] The genetic risk scores corresponding to each tissue are integrated, and the integrated genetic risk score is used as the overall genetic risk score.

[0012] Furthermore, the genetic risk scores corresponding to the tissues are integrated using an equal weight method, a Bayesian model averaging method, or a minimum absolute shrinkage and selection operator to obtain the integrated genetic risk score.

[0013] Furthermore, in S1, according to the marginal effect value β of the j-th SNP j Get the marginal chi-square statistic of the j-th SNP Among them, the marginal effect value β of the j-th SNP j Calculated by the following formula:

[0014] y=X j β j +ε

[0015] In the formula, X j is the n-dimensional genotype vector of the j-th SNP; ε is the n-dimensional residual vector, independent and identically distributed with variance The normal distribution of , and ε f The covariance matrix of

[0016] Furthermore, in S1, according to the marginal effect value β of the j-th SNP j Get the marginal chi-square statistic of the j-th SNP Among them, the marginal effect value β of the j-th SNP j It is estimated by the following formula:

[0017]

[0018] in, Represents β j An estimated value of Represents X j The transpose of X j is the n-dimensional genotype vector of the j-th SNP.

[0019] Further, take Then the b f The optimal linear unbiased estimate of can be simplified to:

[0020]

[0021] Among them, I M represents the M-order unit matrix; R represents the correlation matrix between SNPs obtained from an external data set; It is composed of M The matrix formed by , j is 1,2,…M.

[0022] Furthermore, in S1, the marginal chi-square statistic based on the j-th SNP is The heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested is calculated by the following formula:

[0023]

[0024] in, represents the marginal chi-square statistic of the j-th SNP The expected value of; l(j,c,f) represents the cth cell type in tissue f LD score; Indicates the cell type corresponding to tissue f The heritability of specific annotations.

[0025] According to a second aspect of the present invention, a genetic risk prediction method integrating functional annotation information is provided, comprising:

[0026] Genetic risk prediction is performed using a genetic risk prediction model constructed by the method for constructing a genetic risk prediction model integrating functional annotation information as described in any one of the first aspects.

[0027] According to a third aspect of the present invention, there is provided an electronic device comprising a computer-readable storage medium and a processor;

[0028] The computer-readable storage medium is used to store executable instructions;

[0029] The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the method for constructing a genetic risk prediction model integrating functional annotation information as described in any one of the first aspects, and / or to execute the method for predicting genetic risk integrating functional annotation information as described in the second aspect.

[0030] According to the fourth aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, it implements the method for constructing a genetic risk prediction model integrating functional annotation information as described in any one of the first aspects, and / or implements the genetic risk prediction method integrating functional annotation information as described in the second aspect.

[0031] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0032] (1) The method of the present invention is based on the marginal chi-square statistic of each SNP Calculate the heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested, and then obtain the heritability of the jth SNP in the current tissue f The random effects model is used to estimate the joint effect size b of M SNPs in tissue f f , and based on b f The covariance between the phenotype vector y of the n samples to be tested is b f The optimal linear unbiased estimate of Then we get the genetic risk prediction model In the process of constructing the genetic risk prediction model for each tissue, the heritability of each SNP is used. Instead of the heritability of the entire SNP, the accuracy of genetic risk prediction can be improved when genetic risk prediction is performed using a genetic risk prediction model that is refined based on the heritability of each SNP.

[0033] In addition, the present invention calculates the heritability of specific annotations of each cell type corresponding to each tissue of the sample to be tested, and incorporates the SNP functional annotation information of different tissues into the genetic risk prediction to construct a genetic risk prediction model corresponding to each tissue. Compared with the existing single functional annotation information model, the method of the present invention can integrate the functional annotation information of different tissues for prediction, so as to improve the accuracy of genetic risk prediction.

[0034] (2) Furthermore, the method of the present invention integrates the genetic risk scores corresponding to each tissue, and uses the genetic risk score obtained after integrating the functional annotation information as the genetic risk score of the overall sample to be tested. The overall genetic risk score takes into account the functional annotation information of each tissue, further improving the accuracy of genetic risk prediction. In addition, the overall genetic risk score integrates the genetic risk scores of each tissue, which can further quantify the relative contribution of different tissues to the overall genetic risk of the disease or trait.

[0035] (3) As a preference, the genetic variance (heritability) of a single SNP ) is small, and the phenotypic vector is usually normalized before analysis. Use the M-order identity matrix I M Approximately, that is You can use b f The optimal linear unbiased estimate of Further simplification can improve the computational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Schematic diagram of a method for constructing a genetic risk prediction model in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0038] Example 1

[0039] like Figure 1 As shown, an embodiment of the present invention provides a method for constructing a genetic risk prediction model integrating functional annotation information, which mainly includes:

[0040] S1, marginal chi-square statistic based on the jth SNP Calculate the heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested, and then obtain the heritability of the jth SNP in the current tissue f Wherein, j is 1, 2, ... M, M represents the total number of SNPs in each tissue of the sample to be tested. In the embodiment of the present invention, the sample to be tested is a human body; Indicates the cell type corresponding to tissue f The heritability of the specific annotation, c takes 1, 2, … C, and C represents the total number of cell types corresponding to tissue f.

[0041] S2, using the random effects model y = Xb f +ε f Estimate the joint effect size b of M SNPs in tissue f f ; Where y represents the n-dimensional phenotype vector of the sample to be tested, X is the M*n-dimensional genotype matrix, which is composed of the n-dimensional genotype vectors of M SNPs, ε f represents the n-dimensional residual vector of tissue f, representing the difference between y and the phenotypic prediction value Xb f The difference between them, n is the total number of samples to be tested.

[0042] S3, based on b f The covariance between y and b f The optimal linear unbiased estimate of Among them, D f Indicates b f The covariance matrix of σ 2 Pick and X T represents the transpose of X, Represents ε f The covariance matrix of Determine that I is the identity matrix, and its order is the same as The number of h f It represents the heritability corresponding to tissue f, which is calculated as: l j is the linkage disequilibrium score of the jth SNP, and n represents the number of samples to be tested.

[0043] S4. Constructing a genetic risk prediction model PRS f It represents the genetic risk score corresponding to tissue f. The high or low score is used to reflect the high or low genetic risk.

[0044] Specifically, in S1, according to the marginal effect value β of the j-th SNP j Get the marginal chi-square statistic for the jth SNP Among them, the marginal effect value of the j-th SNP β j By fitting the n-dimensional phenotype vector y of the sample to be tested and the n-dimensional genotype vector X of the j-th SNP j The relationship between them is obtained.

[0045] As a preferred embodiment, the standard linear model in GWAS is used to fit the n-dimensional phenotype vector y of the sample to be tested and the n-dimensional genotype vector X of the j-th SNP. j The relationship between:

[0046] y=X j β j +ε

[0047] Among them, ε is the n-dimensional residual vector, which represents the difference between the n-dimensional phenotype vector y and the phenotype prediction value X. j β j The residuals between are independent and identically distributed with variance The normal distribution of f The covariance matrix of for:

[0048] Usually, the phenotype vector and the genotype vector are standardized to follow a standard normal distribution with a mean of 0 and a variance of 1. Therefore, the marginal effect value β of the j-th SNP is j It can also be estimated by the following formula:

[0049]

[0050] in, Represents β j The estimated value of ; generally the number of samples to be tested n is relatively large.

[0051] Preferably, the sample tissues to be tested in the embodiment of the present invention include adrenal gland and pancreas, central nervous system, cardiovascular, connective tissue and bone, gastrointestinal tract, hematopoietic, kidney, liver, skeletal muscle and other tissues, a total of 10 tissues, and the C cell types corresponding to tissue f are

[0052] Preferably, the stratified LD score regression (S-LDSC) method is used in the embodiment of the present invention, based on the marginal chi-square statistic of the j-th SNP. Calculate the heritability of specific annotations for each cell type corresponding to each tissue in the sample to be tested:

[0053]

[0054] in, represents the marginal chi-square statistic for the j-th SNP The expected value of; l(j,c,f) represents the cth cell type in tissue f LD score, that is, the linkage disequilibrium score between M SNPs; Indicates the cell type corresponding to tissue f The heritability of the specific annotation, that is, the cell type corresponding to tissue f In other embodiments, restricted maximum likelihood estimation (REML) and other algorithms can also be used to calculate

[0055] As a preferred implementation, in S3, b f The covariance between y and y is:

[0056]

[0057] Among them, Cov represents the covariance operator, D f for:

[0058]

[0059] Furthermore, in S3, due to the genetic variance (heritability) of a single SNP ) is small, and the phenotypic vector is usually normalized before analysis. Use the M-order identity matrix I M Approximately, that is According to the Woodbury matrix, b f The optimal linear unbiased estimate of Simplified to:

[0060]

[0061] Where R represents the correlation matrix between SNPs, which is obtained from external datasets, such as 1000 Genomes (1000G) or UK Biobank; For M The matrix formed by , j takes 1, 2, ... M. In this way, the calculation efficiency can be improved.

[0062] Furthermore, the genetic risk prediction model construction method of the present invention also includes:

[0063] The genetic risk scores corresponding to each tissue are integrated, and the integrated genetic risk score is used as the genetic risk score of the overall sample to be tested.

[0064] Preferably, the ensemble includes an Equal Weight (EW), Bayesian Model Averaging (BMA), or Least Absolute Shrinkage and Selection Operator (LASSO) model. Each strategy for adjusting the PRS hyperparameters provides a method for assessing the relative contribution of each cell type to the overall genetic risk of a disease or trait.

[0065] The method of the present invention is based on the marginal chi-square statistic of each SNP Calculate the heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested, and then obtain the heritability of the jth SNP in the current tissue f The random effects model is used to estimate the joint effect size b of M SNPs in tissue f f , and based on b f The covariance between the phenotype vector y of the n samples to be tested is b f The optimal linear unbiased estimate of Then we get the genetic risk prediction model In the process of constructing the genetic risk prediction model for each tissue, the heritability of each SNP is used. Instead of the heritability of the entire SNP, the accuracy of genetic risk prediction can be improved when genetic risk prediction is performed using a genetic risk prediction model that is refined based on the heritability of each SNP.

[0066] As a further design of the present invention, the method of the present invention incorporates the SNP functional annotation information of different tissues into the genetic risk prediction by calculating the heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested, so as to construct a genetic risk prediction model corresponding to each tissue. The genetic risk score corresponding to each tissue is integrated, and the genetic risk score obtained after the integration is used as the genetic risk score of the overall sample to be tested. The overall genetic risk score takes into account the functional annotation information of each tissue, further improving the accuracy of the genetic risk prediction. In addition, the overall genetic risk score incorporates the genetic risk scores of each tissue, which can further quantify the relative contribution of different tissues to the overall genetic risk of the disease or trait.

[0067] In summary, the present invention integrates the PRS of functional annotation information groups of multiple tissues, and provides a comprehensive measure of the overall genetic risk of individual diseases or traits. The method of the present invention helps to construct a PRS with broad applicability and robustness, and can effectively predict the genetic phenotype.

[0068] Example 2

[0069] The embodiment of the present invention provides a genetic risk prediction method integrating functional annotation information, comprising:

[0070] The genetic risk prediction model constructed by the genetic risk prediction model construction method integrating functional annotation information in Example 1 is used to obtain the genetic risk score corresponding to each tissue. The high or low score is used to reflect the high or low genetic risk, thereby completing the genetic risk prediction.

[0071] Furthermore, it also includes integrating the genetic risk scores corresponding to each tissue and using the integrated genetic risk score as the genetic risk score of the overall sample to be tested.

[0072] For the related technical solutions, please refer to the corresponding description in Example 1, which will not be repeated here.

[0073] It should be noted that the above method is completely implemented by computer.

[0074] Example 3

[0075] An embodiment of the present invention provides an electronic device, including a computer-readable storage medium and a processor;

[0076] Computer-readable storage media for storing executable instructions;

[0077] The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the genetic risk prediction model construction method integrating functional annotation information in Example 1, and / or, to execute the genetic risk prediction method integrating functional annotation information in Example 2. For related technical solutions, please refer to the description of the corresponding embodiments, which will not be repeated here.

[0078] Example 4

[0079] The embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the method for constructing a genetic risk prediction model integrating functional annotation information as in Example 1 is implemented, and / or the method for predicting genetic risk integrating functional annotation information as in Example 2 is implemented. For the related technical solutions, please refer to the description of the corresponding embodiments, which will not be repeated here.

[0080] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a genetic risk prediction model integrating functional annotation information, characterized in that: include: S1, marginal chi-square statistic based on the jth SNP Calculate the heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested, and then obtain the heritability of the jth SNP in the current tissue f Where j is 1, 2, ... M, and M represents the total number of SNPs in each tissue of the sample to be tested; Indicates the cell type corresponding to tissue f The heritability of specific annotation, c takes 1, 2, ... C, C represents the total number of cell types corresponding to tissue f; S2, using the random effects model y = Xb f +ε f Estimate the joint effect size b of M SNPs in tissue f f ; where y is the n-dimensional phenotype vector of the sample to be tested, X is the M*n-dimensional genotype matrix, ε f is the n-dimensional residual vector of organization f, n is the total number of samples to be tested; S3, based on b f The covariance between y and b f The optimal linear unbiased estimate of Among them, D f and Respectively represent b f and ε f The covariance matrix of σ 2 Pick u is a constant; S4. Constructing a genetic risk prediction model PRS f It represents the genetic risk score corresponding to tissue f, and the level of the genetic risk score is used to reflect the level of genetic risk.

2. The method for constructing a genetic risk prediction model according to claim 1, characterized in that: After obtaining the genetic risk scores corresponding to each tissue, it also includes: The genetic risk scores corresponding to each tissue are integrated, and the integrated genetic risk score is used as the overall genetic risk score.

3. The method for constructing a genetic risk prediction model according to claim 2, characterized in that: The genetic risk scores corresponding to the tissues are integrated using an equal weight method, a Bayesian model averaging method, or a minimum absolute shrinkage and selection operator to obtain the integrated genetic risk score.

4. The method for constructing a genetic risk prediction model according to claim 1 or 2, characterized in that: In S1, according to the marginal effect value β of the j-th SNP j Get the marginal chi-square statistic of the j-th SNP Among them, the marginal effect value β of the j-th SNP j Calculated by the following formula: y=X j b j +e Where, X j is the n-dimensional genotype vector of the j-th SNP; ε is the n-dimensional residual vector, independent and identically distributed with variance The normal distribution of , and ε f The covariance matrix of 5. The method for constructing a genetic risk prediction model according to claim 1 or 2, characterized in that: In S1, according to the marginal effect value β of the j-th SNP j Get the marginal chi-square statistic of the j-th SNP Among them, the marginal effect value β of the j-th SNP j It is estimated by the following formula: in, Represents β j An estimated value of Represents X j The transpose of X j is the n-dimensional genotype vector of the j-th SNP.

6. The method for constructing a genetic risk prediction model according to claim 5, characterized in that: Pick Then the b f The optimal linear unbiased estimate of can be simplified to: Among them, I M represents the M-order unit matrix; R represents the correlation matrix between SNPs obtained from an external data set; It is composed of M The matrix formed by , j is 1,2,…M.

7. The method for constructing a genetic risk prediction model according to claim 1 or 2, characterized in that: In S1, the marginal chi-square statistic based on the j-th SNP The heritability of the specific annotations of each cell type corresponding to each tissue of the sample to be tested is calculated by the following formula: in, represents the marginal chi-square statistic of the j-th SNP Expected value; represents the cth cell type in tissue f LD score; Indicates the cell type corresponding to tissue f The heritability of specific annotations.

8. A genetic risk prediction method integrating functional annotation information, characterized in that: include: Genetic risk prediction is performed using a genetic risk prediction model constructed by the method for constructing a genetic risk prediction model integrating functional annotation information as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: comprising a computer readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is used to read the executable instructions stored in the computer-readable storage medium to execute the method for constructing a genetic risk prediction model integrating functional annotation information as described in any one of claims 1-7, and / or to execute the method for predicting genetic risk integrating functional annotation information as described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the method for constructing a genetic risk prediction model integrating functional annotation information as described in any one of claims 1 to 7, and / or implements the method for predicting genetic risk integrating functional annotation information as described in claim 8.

Images