Multi-gene risk scoring model construction method and system based on multi-population multi-phenotype data integration

Abstract

The invention discloses a multi-gene risk scoring model construction method and system based on multi-population multi-phenotype data integration, and the method comprises the steps: S1, carrying out the construction of a to-be-detected phenotype according to a to-be-detected phenotype list composed of a to-be-detected phenotype and other phenotypes having genetic correlation with the to-be-detected phenotype; collecting phenotype GWAS data and reference genome data required for analysis in the descendant crowd to which the sample to be detected belongs and other descendant crowds, and collecting genotype data of the sample to be detected; s2, respectively calculating an LD score between every two sites and a comprehensive LD score of each site; s3, estimating parameters of the polygene risk scoring model, including the genetic rate of each phenotype, the genetic covariance and the environmental covariance between every two phenotypes, and the cross-population genetic covariance of each phenotype of the descendant population to which the sample to be tested belongs and other descendant population, and estimating a site weight value of a polygene risk score PRS; and S4, calculating a polygene risk score PRS value of the to-be-detected sample. The prediction precision of the complex phenotype or disease risk is improved.

Description

Technical Field

[0001] The present invention relates to a polygenic risk scoring model, and in particular to a polygenic risk scoring model construction method and system based on multi-population multi-phenotype data integration, belonging to the application field of combining biological genetic data analysis with computers. Background Art

[0002] Studies have shown that most complex phenotypes are affected by many genetic variants with small effect sizes, which is called "polygenicity". Polygenic risk scores (PRS) constructed based on SNPs collected across the genome in genome-wide association studies (GWAS) have shown great potential in personal and clinical applications for a variety of heritable diseases, including assisting diagnosis, early and cost-effective intervention, and improved treatment strategies by stratifying patients into different risk groups. For example, a recent PRS model for type 2 diabetes (T2D) based on 600,000 European samples achieved an area under the receiver operating characteristic curve (AUC) of 0.65 on the test set. Among them, individuals in the top 5% of the PRS had a three-fold increased risk of T2D compared to the average individual. However, since 94% of the GWAS samples to date are from European populations, the early constructed PRS tends to have better predictive power in European populations. When the type 2 diabetes (T2D) PRS model was applied to individuals of African descent, the AUC of the PRS model dropped to 0.57, indicating that the applicability of PRS among different populations is very limited.

[0003] In recent years, more and more scientists have begun to improve the predictive ability of PRS in non-European populations by integrating information from large-scale European population datasets. Examples of the methods used are as follows: (1) MultiPRS linearly combines PRS models from multiple populations using mixed weights estimated by cross-validation. (2) Linear mixed models (LMMs) are widely used in genome-wide association studies, and their estimated SNP effects are called best unbiased linear predictors (BLUPs) under the assumption of random effects. (3) XP-BLUP, based on BLUP, assumes that significant SNPs identified in large-scale European populations also play an important role in constructing PRS for non-European populations. Therefore, it uses additional parts in LMM to characterize significant SNPs pre-selected from large-scale GWAS of European populations. (4) XPA, through novel data structure and algorithm design, can construct PRS for non-European populations using biobank-scale data. (5) PRS-CSx jointly models GWAS summary statistics data from multiple populations by introducing continuous shrinkage priors. Therefore, PRS-CSx can model the shared genetic effects of causal variants across populations while allowing SNP effect sizes to vary between populations. Despite the above progress in constructing PRS across populations, the above methods mainly focus on a single phenotype.

[0004] To expand, genetic variants often affect multiple phenotypes, which is called genetic pleiotropy. In a systematic analysis of 4,155 publicly available GWAS, 90% of phenotype-associated loci showed genetic pleiotropy. These loci may simultaneously image multiple phenotypes and induce correlations between them. In fact, genetic correlations estimated from GWAS also reveal the prevalence of pleiotropy in complex human phenotypes or diseases. Some classic examples include the strong genetic correlation between bipolar disorder and schizophrenia (0.68, standard error = 0.04), and the moderate genetic correlation between T2D and body mass index (BMI) (0.4, standard error = 0.04). Therefore, by exploiting the shared genetic basis, constructing PRS across multiple related phenotypes can further improve risk prediction. In fact, in PRS analysis, the advantages of multiple phenotypes over single phenotypes have been fully verified, and PRS methods including MTGBLUP, PleioPred and SMTpred have begun to try to integrate GWAS data of multiple phenotypes to improve the predictive power of the target phenotype. However, these PRS methods mainly focus on GWAS datasets of a single European population. There is still no method that can effectively integrate genetic pleiotropy among multiple phenotypes while taking advantage of large-scale European datasets to improve the predictive ability of PRS models for target phenotypes in the target population.

[0005] In summary, the existing PRS method still has shortcomings in simultaneously integrating multi-population and multi-phenotype data. This patent proposes a new method for constructing a PRS model based on the integration of multi-population and multi-phenotype data. Summary of the invention

[0006] In response to the above-mentioned existing technical problems, the present invention provides a PRS model construction method and system based on the integration of multi-population and multi-phenotype data, which integrates multi-population and multi-phenotype data to achieve the technical purpose of improving the accuracy of specific PRS, thereby improving the prediction accuracy of complex phenotypes or disease risks.

[0007] To achieve the above object, firstly, the present invention provides a method for constructing a PRS model based on the integration of multi-population and multi-phenotype data, comprising the following steps: S1. According to the list of phenotypes to be tested consisting of the phenotypes to be tested and other phenotypes with genetic correlation with them, collect the phenotype GWAS data required for analysis in the ethnic population to which the sample to be tested belongs and other ethnic populations, as well as the reference genome data, and collect the genotype data of the sample to be tested; S2, based on the reference genome data collected in S1, calculate the LD score between every two sites and the comprehensive LD score of each site; S3. Based on the comprehensive LD score obtained in S2 and the phenotypic GWAS data collected in S1, estimate the parameters of the polygenic risk score model, including the heritability of each phenotype in the ethnic population to which the sample to be tested belongs and other ethnic populations, the genetic covariance and environmental covariance between every two phenotypes, and the genetic covariance of each phenotype across populations, and estimate the site weight value of the polygenic risk score PRS; S4. Calculate the polygenic risk score PRS value of the sample to be tested based on the genotype data collected in S1 and the site weight value obtained in S3.

[0008] Specifically, the calculation formula of the LD score in S2 is as follows: in, , Respectively represent The site, individual sites; is the LD score of the two loci; , It is the first of two sites The genotypes of the samples; , is the average of the genotypes at the two loci.

[0009] The method of the present invention is further provided that the calculation formula of the comprehensive LD score in S2 is as follows: in, It is The comprehensive LD score of each locus; , They are The site, individual sites; is the LD score between the two loci.

[0010] Specifically, the method of the present invention is as follows: in, It is The heritability of a phenotype; is the number of sites; It is Sample size for each phenotype; and, It is an estimate of the regression coefficient obtained by regressing the square z-score of the site in the GWAS data of the same population onto the LD score of the site. The calculation formula is as follows: in, It is The phenotype of GWAS z-score of each locus; It is The comprehensive LD score of each locus; It is an estimate of the intercept obtained by regressing the squared z-score of a site in GWAS data within the same population onto the LD score of the site.

[0011] Specifically, the method of the present invention is as follows: in, is the estimate of the genetic covariance between the first and second phenotypes; is the sample size of the first phenotype; is the sample size of the second phenotype; is the number of sites; and, It is the estimated value of the regression coefficient obtained by regressing the product of the z-score of each site in the GWAS data of the first phenotype in the ethnic population to which the sample to be tested belongs and the first phenotype in other ethnic populations onto the site LD score. The calculation formula is as follows: in, , is the first order of magnitude of the same phenotype in the ethnic group to which the sample belongs and in other ethnic groups. GWAS z-score of each locus; It is The comprehensive LD score of each locus; are the estimated values ​​of the regression coefficient and intercept, respectively.

[0012] Specifically, the method of the present invention is as follows: in, is the estimate of the environmental covariance between the first and second phenotypes; It is an estimate of the intercept obtained by regressing the z-score of the site in the GWAS data within the same population onto the LD score of the site; is the sample size of the first phenotype; is the sample size of the second phenotype; is the number of repeated individuals in the two phenotype GWAS data; is an estimate of the genetic covariance between the first and second phenotypes.

[0013] Specifically, the method of the present invention is as follows: in, Is the first among the ethnic group of the sample to be tested and other ethnic groups Estimates of the cross-population genetic covariance of phenotypes; is the number of sites; It is The sample size of each phenotype in the ethnic group to which the sample to be tested belongs; It is The sample size of each phenotype in other ethnic groups; and, The first phenotypes and the first The estimated value of the regression coefficient obtained by regressing the product of the z-score of each site in the GWAS data of the phenotype to the site LD score is calculated as follows: in, , Is the first among the ethnic group of the sample to be tested and other ethnic groups The respective GWAS z-score of each locus; It is The comprehensive LD score of each locus; The first phenotypes and the first The intercept is estimated by regressing the product of the z-score of each locus in the GWAS data for the phenotype onto the locus LD score.

[0014] Specifically, the method of the present invention is as follows: in, It is the first Phenotype The posterior mean of the estimated genetic effect values ​​for each locus; It is the first Phenotype The posterior mean of the estimated genetic effect values ​​for each locus; It is the first Phenotype The specific genetic effect value of each locus; It is the first Phenotype The specific genetic effect value of each locus; It is Phenotype The standard deviation of the genotype at each locus; is the number of sites; and, and The calculation method is as follows: in, is the locus-specific genetic effect value; is the genotype matrix of loci with population- and phenotype-specific genetic effects; is the phenotypic value after removing the influence of covariates; yes The inverse of the marginal distribution covariance matrix; is the posterior mean vector of the estimated random genetic effect values ​​of the loci; X is the genotype matrix; the symbol represents the Kronecker product; is the number of sites, yes dimensional identity matrix; is the inverse of the covariance matrix of the site random effect; is the inverse matrix of the environmental covariance matrix; and, , and The calculation method is as follows: in, yes Marginal distribution covariance matrix; yes The inverse of the marginal distribution covariance matrix; is the inverse of the covariance matrix of the site random effect; is the inverse of the environmental covariance matrix; X is the genotype matrix; the symbol represents the Kronecker product; is the number of sites, yes dimensional identity matrix, , , The subscripts correspond to the identity matrix of the dimension respectively; It is The heritability of a phenotype; It is The phenotype and The genetic covariance of the phenotypes; is the number of individuals overlapping in the two studies; It is The phenotype and Environmental covariance of phenotypes; and, Approximately: Among them, the symbol and denote element-wise product and Kronecker product respectively; and is the LD matrix of all sites; and It is the LD matrix between all loci and large effect loci in the ethnic group to which the sample to be tested belongs and other ethnic groups; matrix and matrix are the standardized genotype matrices from the reference genomes of the ethnic group to which the sample to be tested belongs and other ethnic groups, with a mean of 0 and a variance of ; z is the z-score of the GWAS genetic effect value; are the GWAS sample sizes of phenotype 1 and phenotype 2 in the target population and the population to be tested, respectively. is the number of overlapping individuals in two GWAS studies of the same population.

[0015] Specifically, the method of the present invention is as follows: in, For new individuals or samples, PRS value under each phenotype; It represents the genotype data of a new individual, which contains the codes of multiple SNPs; is a vector, which refers to the estimates of the random genetic effects of all loci for a phenotype; It corresponds to the first A genotype data matrix for specific effect sites of each phenotype; is a vector containing Estimates of specific effect sizes associated with subsets of SNPs.

[0016] In addition, the present invention also provides a polygenic risk scoring model construction system based on the integration of multi-population and multi-phenotype data, including a data entry and storage module, a data preprocessing module, a parameter estimation module, and a PRS calculation module; The data entry and storage module is used to collect phenotype GWAS data required for analysis from the ethnic group to which the sample to be tested belongs and other ethnic groups, as well as reference genome data, and collect genotype data of the sample to be tested, based on a list of phenotypes to be tested consisting of the phenotypes to be tested and other phenotypes genetically correlated therewith; The data preprocessing module calculates the LD score between two loci and the comprehensive LD score of each locus according to the collected reference genome data; The parameter estimation module is used to estimate the parameters of the polygenic risk score model based on the obtained comprehensive LD score and the collected phenotype GWAS data, including the heritability of each phenotype of the ethnic population to which the sample to be tested belongs and other ethnic populations, the genetic covariance and environmental covariance between every two phenotypes, and the genetic covariance of each phenotype across populations, and estimate the site weight value of the polygenic risk score PRS; The PRS calculation module is used to calculate the polygenic risk score PRS value of the sample to be tested based on the collected genotype data and the obtained site weight value.

[0017] In summary, the present invention constructs a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data, which has the following beneficial effects: first, other phenotypes related to the phenotype to be tested are taken into account, and the PRS weight is optimized based on the genetic correlation between different phenotypes, thereby improving the prediction accuracy of the PRS model; secondly, the vast majority of currently publicly available sample data come from European populations, while data from Asian populations are relatively scarce. Therefore, the present invention is based on the genetic correlation between different populations and utilizes a large-scale European population data set to improve the prediction accuracy of a small number of populations.

[0018] Compared with the prior art, the present invention mainly has the following technical advantages: 1) Improve the prediction accuracy of the PRS model: Compared with the existing PRS technology, the present invention significantly improves the prediction accuracy of the PRS model by integrating data from multiple populations and multiple phenotypes.

[0019] 2) Enhanced cross-population applicability: The applicability of the PRS model of the present invention in different populations is significantly improved, especially by utilizing large-scale European population data to improve the predictive ability of PRS for non-European populations.

[0020] 3) Dealing with the pleiotropy of genetic effects: Compared with the existing PRS model, the present invention more effectively deals with the pleiotropy of genetic effects, explores the genetic correlations between multiple phenotypes, and improves the comprehensive utility of the PRS model.

[0021] 4) Personalized risk assessment: The present invention allows for more personalized risk assessment of target populations and target phenotypes, and further improves the accuracy of genetic risk prediction by combining significant genetic effects that are phenotype-specific and population-specific. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A schematic diagram of a process provided for an embodiment of the method of the present invention; Figure 2 A comparison chart of the prediction performance of the PRS calculation method provided in the embodiment of the method of the present invention and other methods. DETAILED DESCRIPTION

[0024] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention. Unless otherwise specified in the specification, all scientific and technical terms used in the specification have the same meanings as those generally understood by those skilled in the art to which the present invention belongs.

[0025] It should be noted that the "multiple populations" in the present invention are usually but not limited to two populations. When the present invention is applied to two populations: one is the population to which the sample to be tested belongs (also referred to as the target population), and the sample to be tested refers to the sample whose polygenic risk score (PRS) needs to be calculated; the other is the auxiliary population. Specifically, on the one hand, the main difference between the population to which the sample to be tested belongs and the auxiliary population is the different ethnicity, for example: East Asian population and European population, African population and European population. On the other hand, the population to which the sample to be tested belongs and the auxiliary population can be any population, that is, the model constructed by the present invention can be applied to any combination of populations. However, in practice, since the European population data is the most in the GWAS data published worldwide, and other populations such as East Asian population or African population are relatively small, when studying East Asian population or African population, the European population is usually used as an auxiliary population, so that the advantage of the large amount of data of the European population can be used to improve the analysis results of the small amount of data in the East Asian population and the African population. In addition, when the present invention is applied to three or more groups of people: one of the groups is the group to which the sample to be tested belongs, and the other groups are auxiliary groups. Example 2 describes the extended application method of the present invention to three or more groups of people.

[0026] It should be noted that the "multiple phenotypes" in the present invention mainly involve two types of phenotypes: one is the phenotype for which a polygenic risk score (PRS) is to be calculated (hereinafter referred to as phenotype 1), such as type 2 diabetes (T2D); the other is other phenotypes that are genetically correlated with phenotype 1 (hereinafter referred to as phenotype 2), for example, there are multiple other phenotypes that are genetically correlated with T2D, which may be body mass index (BMI), total cholesterol (TC), triglycerides (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), and a list of phenotypes to be tested is obtained based on phenotype 1 and phenotype 2. In other embodiments, other phenotypes 1 and phenotype 2 may be selected as needed, and the number of phenotypes 2 is not limited, but the data used are different, and the calculation steps are exactly the same, and the extended application method of the present invention in multiple phenotypes is described in Example 3.

[0027] It should be noted that in the existing technology, genome-wide association analysis (GWAS) is an important means widely used to find genes associated with complex genetic diseases. By studying the association between variant sites on chromosomes and diseases or other traits, analyzing the differences in allele frequencies between diseased and non-disease-affected people, or between people with different traits, the estimated genetic effect of the site is obtained, and the most critical sites affecting the disease or complex traits are screened out through statistical significance (P-value). Here, GWAS data mainly refers to the estimated genetic effect or Z-score of the site, as well as the corresponding P-value.

[0028] In addition, the whole genome association study (GWAS) uses a logistic regression model, with a specific phenotype as the dependent variable, to model and calculate each gene locus separately. In addition to genomic data, the model usually adds age, gender, the first five principal components PC and data batch as covariates, so as to obtain an estimate of the genetic effect value of each locus on the phenotype. In addition, the genetic effect value measures the degree of genetic explanation of the variation of the locus for the phenotype.

[0029] Example 1: A method for constructing a polygenic risk scoring model for type 2 diabetes (T2D) based on the integration of two phenotype data from two populations.

[0030] like Figure 1 As shown, this embodiment provides a method for constructing a polygenic risk scoring model based on the integration of multi-population multi-phenotype data, specifically, a method for constructing a polygenic risk scoring model for type 2 diabetes (T2D) based on the integration of two population two-phenotype data, which is introduced as follows.

[0031] S1. Based on the list of phenotypes to be tested and other phenotypes that are genetically related to the phenotypes to be tested, collect the phenotype GWAS data required for analysis in the ethnic population to which the sample to be tested belongs and other ethnic populations, as well as the reference genome data, and collect the genotype data of the sample to be tested.

[0032] It should be noted that, in this embodiment, the population to which the sample to be tested belongs and the auxiliary population are respectively from the East Asian population to which the sample to be tested belongs and the European population, and PRS calculation is performed based on type 2 diabetes (T2D) in the East Asian population and the European population.

[0033] In specific implementation, this step S1 includes the following sub-steps: S1-1. Based on the list of phenotypes to be measured, including body mass index (BMI), total cholesterol (TC), triglycerides (TG), high-density lipoprotein (HDL) and low-density lipoprotein (LDL) that are genetically correlated with type 2 diabetes (T2D), the phenotypic GWAS data required for analysis in East Asian and European populations were collected and used as training data sets.

[0034] In the specific implementation, first, since the Biobank of Japan (BBJ, sample size 108,479) and the UK Biobank (UKBB, sample size 459,324) are two of the largest and most commonly used GWAS databases in the world, this step uses BBJ and UKBB to represent East Asian populations and European populations, respectively, and collects the type 2 diabetes (T2D) genome-wide association study (GWAS) summary statistics data from BBJ and UKBB to obtain the training data set. In addition, "-EAS" or "-EUR" is added after the phenotype name abbreviation to indicate the GWAS summary statistics data from BBJ or UKBB.

[0035] Secondly, this step takes type 2 diabetes (T2D) as phenotype 1 and the six phenotypes associated with T2D as phenotype 2. The GWAS data of phenotype 1 are collected, including the GWAS data of type 2 diabetes (T2D) from BBJ and UKBB, and the GWAS data of phenotype 2 are collected, including the GWAS data of body mass index (BMI), total cholesterol (TC), triglycerides (TG), high-density lipoprotein (HDL) and low-density lipoprotein (LDL) from BBJ, and the GWAS data of body mass index (BMI) from UKBB.

[0036] Furthermore, GWAS data is usually presented in the form of a data table about site information, and its fields include the site ID, coordinate information, and corresponding genetic effect value, standard deviation of genetic effect value, statistical significance (P value) and other information corresponding to each site. In this step, GWAS data mainly refers to the estimated genetic effect value or Z-score of the site, and the corresponding P-value.

[0037] In addition, the phenotype GWAS data required for the analysis collected in this step are shown in Table 1. This embodiment only selected GWAS data of two groups corresponding to T2D and BMI to incorporate group information into the model, and then used several other phenotypes related to T2D to further enhance the statistical performance of the final T2D PRS.

[0038] It should be noted that there is no GWAS data for TC, TG, HDL, and LDL in the EUR population in Table 1 below. Theoretically, these data can be collected, but in application, it is not necessary to have GWAS data for multiple populations for each phenotype. GWAS data can be collected according to analysis needs, usually GWAS data for the phenotype to be tested in multiple populations, as well as GWAS data for auxiliary phenotypes.

[0039] Table 1 S1-2. Collect reference genome data of East Asian and European populations.

[0040] It should be noted that the reference genome data in this step is public data of human genome data, which is a genome sequence widely used for comparison and analysis of genome variation, such as the human genome data of the 1000 Genome Project. In the prior art, the 1000 Genome Project (1000G database) is an international public project that aims to establish the most detailed and representative human genetic variation database.

[0041] In specific implementation, genome data of East Asian and European populations in the 1000 Genome project will be collected.

[0042] S1-3. Collect genotype data of East Asian samples to be tested.

[0043] It should be noted that the genotype data of the sample to be tested in this step refers to the genotype data of the sample for which the polygenic risk score (PRS) is to be calculated. In addition, for the target population in East Asia, the chip typing data can be obtained by gene chip technology or high-throughput sequencing technology, and Minimac4 can be used for gene filling to form the genome data of the sample to be tested together with the chip typing data.

[0044] In the specific implementation, genotype data of 4,367 East Asian samples (also samples to be tested) in the Genetic Epidemiology of Health and Aging in Adults (GERA) cohort were collected and used as test data for the polygenic risk scoring model of this embodiment, so that the weight of the PRS can be calculated using the GWAS data first, and then the PRS of the sample to be tested can be obtained using the obtained weight and the genotype data of the sample to be tested.

[0045] S2. Based on the reference genome data of the ethnic group to which the sample to be tested belongs and other ethnic groups collected in step S1, calculate the LD score between every two sites and the comprehensive LD score of each site.

[0046] In specific implementation, this step S2 includes the following sub-steps: S2-1. Based on the reference genome data of East Asian and European populations collected in step S1-2, the LD scores between two loci are calculated for the reference genome data.

[0047] It should be noted that in the prior art, the LD score (Linkage Disequilibrium Score) is used to measure the deviation between the frequency of co-occurrence of alleles at two or more gene loci in a population and the frequency of random combination. The commonly used calculation method is the square of the correlation coefficient between the allele frequencies of the two gene loci, and its value is between 0 and 1. The higher the value, the stronger the LD between the two gene loci.

[0048] The calculation formula of the LD score is as follows: in, , Respectively represent The site, individual sites; is the LD score of the two loci; , It is the first of two sites The genotype of each sample; , is the average of the genotypes at the two loci.

[0049] In specific implementation, the LD score calculation formula is used to calculate the LD score between two loci for the reference genome data of East Asian populations and European populations, and the LD score is obtained. value.

[0050] S2-2, based on the LD score calculated in step S2-1, that is, The comprehensive LD score of each site was calculated for the reference genome data of East Asian and European populations.

[0051] It should be noted that the calculation formula of the comprehensive LD score is as follows: in, It is The comprehensive LD score of each locus; , Respectively represent The site, individual sites; is the LD score between the two loci.

[0052] In specific implementation, the calculation formula of the comprehensive LD score is used to calculate the comprehensive LD score of each site for the reference genome data of the East Asian population and the European population. In addition, the entire reference genome data of the East Asian population and the European population collected in step S1-2 is divided into thousands of nearly independent LD blocks (for example, the East Asian population has 1,445 LD blocks), and then the comprehensive LD score of each site is calculated only in a given block LD matrix each time.

[0053] S3. Based on the comprehensive LD score obtained in step S2 and the phenotypic GWAS data of the ethnic group to which the sample to be tested belongs and other ethnic groups collected in step S1, estimate the parameters of the polygenic risk score model, including the heritability of each phenotype of the ethnic group to which the sample to be tested belongs and other ethnic groups, the genetic covariance and environmental covariance of the phenotype to be tested and each other phenotype, and the genetic covariance of each phenotype across populations, and estimate the site weight of the polygenic risk score PRS.

[0054] In specific implementation, this step S3 includes the following sub-steps: S3-1. Based on the comprehensive LD score of each site in the reference genome data obtained in step S2 and the phenotypic GWAS data of East Asian and European populations collected in step S1, estimate the heritability of each phenotype in East Asian and European populations respectively.

[0055] It should be noted that, first of all, in the existing LD score regression (LDSC) model, by regressing the z-score (a standardized indicator reflecting the strength of association between SNP and disease or trait) of GWAS (genome-wide association study) analysis of SNP (single nucleotide polymorphism) onto the LD score (an indicator measuring the degree of linkage disequilibrium of SNP in the genome), an estimated value of heritability can be obtained. This relationship is precisely given by the following formula: in, is the mathematical expectation; It is The phenotype of GWAS z-score of each locus; It is Sample size for each phenotype; It is The heritability of a phenotype; is the number of sites; For the The comprehensive LD score of each locus; is the expansion constant due to confounding bias.

[0056] In addition, the above calculation formula uses the z-score of the GWAS data site, while the publicly published GWAS data usually only include information such as the genetic effect value of the site and the standard deviation of the genetic effect value. Therefore, it is necessary to first divide the genetic effect value by the standard deviation of the genetic effect value to obtain the z-score of the site, and then bring the z-score of the site into the calculation formula to calculate the heritability of the phenotype.

[0057] Based on the above formula, a regression model is established by combining the square z-score of the locus with the locus LD score to estimate the heritability, and its general form can be expressed as: in, It is The phenotype of GWAS z-score of each locus; For the The comprehensive LD score of each locus; It is an estimate of the intercept obtained by regressing the squared z-score of a site in GWAS data within the same population onto the LD score of the site.

[0058] Specifically, for each SNP, there is its z-score and corresponding LD score in GWAS. Through linear regression, a best fit line can be found that minimizes the residual sum of squares between the squared z-score and the LD score. The slope of this best fit line is the estimated value of the regression coefficient, that is, .pass Along with other parameters, estimates of heritability can be calculated, helping to understand how genetic variation at different loci is associated with a disease or trait and how this association is affected by linkage disequilibrium.

[0059] Based on the above regression model, the calculation formula for the heritability of each phenotype is as follows: in, It is The heritability of a phenotype; is the estimated value of the regression coefficient; is the number of sites; It is Sample size for each phenotype.

[0060] In the specific implementation, step S1 collects eight groups of GWAS data, namely, T2D-EAS, T2D-EUR, BMI-EAS, BMI-EUR, TC-EAS, TG-EAS, HDL-EAS, and LDL-EAS, wherein EAS and EUR represent East Asian populations and European populations, respectively. Now these GWAS data are respectively brought into the calculation formula of the heritability of the phenotype, and the heritability of each phenotype corresponding to the eight groups of GWAS data can be calculated.

[0061] In addition, the two phenotypes of East Asian and European populations are represented by subscripts. Indicates phenotype 1 and phenotype 2 of the target population, subscript , Representing phenotype 1 and phenotype 2 of the auxiliary population, regression models are established for the two phenotypes of the two populations respectively, and the phenotypic heritability obtained includes: The above is an example of two populations and two phenotypes, which can also be extended to multiple populations and multiple phenotypes, and different subscripts can be used to represent different phenotypes.

[0062] S3-2. Estimate the genetic covariance and environmental covariance between each two phenotypes in East Asian and European populations, respectively.

[0063] It should be noted that genetic covariance refers to the correlation between genetic variables of two or more traits in genetics. It reflects the degree of genetic association between different phenotypes. Similarly, environmental covariance reflects the correlation of the effects of environmental factors on different phenotypes. In addition, the genetic covariance and environmental covariance of any two phenotypes in East Asian and European populations can be calculated using the following method.

[0064] First, the genetic covariance between each two phenotypes can be estimated according to the following relationship: in, The first phenotype GWAS z-score of each locus; It is the second phenotype GWAS z-score of each locus; is the sample size of the first phenotype; is the sample size of the second phenotype; is the genetic covariance of the two phenotypes; is the number of sites; For the The comprehensive LD score of each locus; is the number of repeated individuals in the two phenotype GWAS data; is the phenotypic correlation in replicate samples.

[0065] Furthermore, the product of the z-score of each locus in the GWAS data of the two phenotypes was regressed onto the locus LD score, resulting in the following regression model: in, The first phenotype GWAS z-score of each locus; is the GWAS z-score of the jth site of the second phenotype; For the The comprehensive LD score of each locus; are the estimated values ​​of the regression coefficient and intercept, respectively.

[0066] The calculation formula for the estimated value of the genetic covariance between each two phenotypes is as follows: in, is the estimate of the genetic covariance between the first and second phenotypes; is the estimated value of the regression coefficient; is the number of sites; is the sample size of the first phenotype; is the sample size of the second phenotype.

[0067] The calculation formula of the estimated value of the environmental covariance is as follows: in, is the estimate of the environmental covariance between the first and second phenotypes; is the sample size of the first phenotype; is the sample size of the second phenotype; is the number of repeated individuals in the two phenotype GWAS data; is an estimate of the genetic covariance between phenotype 1 and phenotype 2. In particular, and It is an estimate of the regression coefficient and intercept obtained by regressing the product of the z-score of each locus in the GWAS data of two phenotypes onto the locus LD score.

[0068] In specific implementation, this step calculates the genetic covariance and environmental covariance of T2D with each other phenotype for the East Asian population (EAS) and the European population (EUR), including the genetic covariance and environmental covariance of T2D with BMI, T2D with TC, T2D with TG, T2D with HDL, and T2D with LDL in the East Asian population (EAS), and the genetic covariance and environmental covariance of T2D with BMI in the European population (EUR).

[0069] In addition, the two phenotypes of East Asian and European populations are represented by subscripts. Indicates phenotype 1 and phenotype 2 of the target population, subscript represents phenotype 1 and phenotype 2 of the auxiliary population, so regression models are established for the two phenotypes of the two populations respectively, and the genetic covariance is obtained and the environmental covariance The above is an example of two populations, which can be extended to multiple populations, and different subscripts can be used to represent the phenotypes of different populations.

[0070] S3-3. Estimate the cross-population genetic covariance of each phenotype in East Asian and European populations respectively.

[0071] It should be noted that cross-population genetic covariance refers to the genetic covariance of the same phenotype in different populations, namely the East Asian population (EAS) and the European population (EUR). In addition, the cross-population genetic covariance estimation method in this step is basically the same as the method in step S3-2, except that the data selected when establishing the regression model is a combination of phenotypes in two populations.

[0072] The calculation formula of the genetic covariance across populations is as follows: in, It is the estimated value of the cross-population genetic covariance of the first phenotype in the ethnic group to which the sample to be tested belongs and in other ethnic groups; is the estimated value of the regression coefficient; is the number of sites; is the sample size of the first phenotype; is the sample size of the second phenotype.

[0073] and, The product of the z-score of each site in the GWAS data of the first phenotype in the ethnic group to which the sample to be tested belongs and the first phenotype in other ethnic groups is regressed to the site LD score: in, , is the first order of magnitude of the same phenotype in the ethnic group to which the sample belongs and in other ethnic groups. GWAS z-score of each locus, For the The comprehensive LD score of each locus; are the estimated values ​​of the regression coefficient and intercept, respectively.

[0074] Similarly, , , The estimation method and Exactly the same, just change , Replaced by a combination of GWAS z-scores of different phenotypes in the two populations.

[0075] In specific implementation, this step calculates the cross-population genetic covariance of the two phenotypes T2D and BMI for the two populations EAS and EUR, respectively, and obtains a cross-population genetic covariance of 0.678 for T2D-EUR and T2D-EAS, and a cross-population genetic covariance of 0.555 for BMI-EUR and BMI-EAS. In other embodiments, the cross-population genetic covariance of multiple phenotypes in multiple populations can be calculated according to the above method.

[0076] S3-4. Estimate the site weight value of the polygenic risk score PRS.

[0077] It should be noted that in practical applications, PRS (polygenic risk score) has been widely used in genetic risk prediction, disease classification and genetic epidemiology research. The existing PRS is calculated based on the individual's genotype data and the effect size (or weight) obtained from the GWAS study. In addition, the sites are sorted according to the GWAS P-value, and the P-value threshold is selected. The sites with a P-value less than the threshold are recorded as large effect site groups (hereinafter referred to as large effect sites).

[0078] Specifically, the calculation model of PRS is: in, For new individuals or samples, PRS value under each phenotype; It represents the genotype data of a new individual, which contains the codes of multiple SNPs; is a vector, which refers to the estimates of the random genetic effects of all loci for a phenotype; It corresponds to the first A genotype data matrix for specific effect sites of each phenotype; is a vector containing Estimates of specific effect sizes associated with subsets of SNPs.

[0079] Specifically, For the Each phenotype has population-specific and phenotype-specific SNP genotype data; yes A subset of SNPs is usually selected based on the SNPs with significant genetic effects in the population. These SNPs are selected based on the p-values ​​obtained from the GWAS analysis of their corresponding phenotypes, which can be obtained from public GWAS statistics. For example, the p-value in the GWAS of the first phenotype in the target population is less than a given threshold (e.g. ), then these SNPs are selected as the specific SNPs of the first phenotype of the target population, and the corresponding genotype data are Similarly, the same selection process is applied to other phenotypes of the target population as well as the auxiliary population.

[0080] and, and The calculation formula is as follows: in, It is the first Phenotype The posterior mean of the estimated genetic effect values ​​for each locus; It is the first Phenotype The posterior mean of the estimated genetic effect values ​​for each locus; It is the first Phenotype The specific genetic effect value (or fixed genetic effect value) of each locus; It is the first Phenotype The specific genetic effect value of each locus; It is Phenotype The standard deviation of the genotype at each locus; is the number of sites.

[0081] and, and The calculation formula is as follows: in, is the locus-specific genetic effect value; is the genotype matrix of loci with population- and phenotype-specific genetic effects; is the phenotypic value after removing the influence of covariates; yes The inverse of the marginal distribution covariance matrix; is the posterior mean vector of the estimated random genetic effect values ​​of the loci; X is the genotype matrix; the symbol represents the Kronecker product; is the number of sites, yes dimensional identity matrix; is the inverse of the covariance matrix of the site random effect; is the inverse of the environmental covariance matrix.

[0082] and, , and The calculation method is as follows: in, yes Marginal distribution covariance matrix; yes The inverse of the marginal distribution covariance matrix; is the inverse of the covariance matrix of the site random effect; is the inverse of the environmental covariance matrix; X is the genotype matrix; the symbol represents the Kronecker product; is the number of sites, yes dimensional identity matrix; , , The subscripts correspond to the identity matrix of the dimension respectively; It is The heritability of a phenotype; It is The phenotype and The genetic covariance of the phenotypes; is the number of individuals overlapping in the two studies; It is The phenotype and Environmental covariance of a phenotype.

[0083] The above formula requires individual-level genetic data for calculation. The present invention calculates by using GWAS summary statistics data and LD reference data and An approximate solution of It can be approximated as: Among them, the symbol and denote element-wise product and Kronecker product respectively; and is the LD matrix of all sites; and is the LD matrix between all loci and large-effect loci in the two populations; matrix and matrix are the standardized genotype matrices (with a mean of 0 and a variance of ); z is the z-score of the GWAS genetic effect value, are the GWAS sample sizes of phenotype 1 and phenotype 2 in the target population and the population to be tested, respectively. is the number of overlapping individuals in two GWAS studies of the same population.

[0084] In the specific implementation, the heritability of each phenotype in the East Asian population and the European population obtained in steps S3-1, S3-2, and S3-3, the genetic covariance and environmental covariance of the two phenotypes, and the genetic covariance of each phenotype across populations are first introduced into and The calculation formula is then based on and The calculation formula is used to calculate the site weight and The estimated value of .

[0085] In addition, the entire reference genome collected in step S1 is divided into thousands of nearly independent LD blocks (for example, East Asian populations have 1,445 LD blocks), and then the site weights are calculated only within a given block LD matrix each time. and The estimated value of .

[0086] S4. Calculate the polygenic risk score PRS value of the sample to be tested based on the genotype data of the sample to be tested collected in step S1 and the site weight value obtained in step S3.

[0087] It should be noted that the calculation formula of the polygenic risk score PRS value is as follows: in, For new individuals or samples, PRS value under each phenotype; It represents the genotype data of a new individual, which contains the codes of multiple SNPs; is a vector, which refers to the estimates of the random genetic effects of all loci for a phenotype; It corresponds to the first A genotype data matrix for specific effect sites of each phenotype; is a vector containing Estimates of specific effect sizes associated with subsets of SNPs.

[0088] In specific implementation, the genotype data of the East Asian sample to be tested and the site weight of the polygenic risk score PRS obtained in step S3 are brought into the calculation formula of the polygenic risk score PRS value to obtain the polygenic risk score PRS value of the East Asian sample to be tested.

[0089] at last, Figure 2 The fitting performance of the PRS for T2D obtained by the test sample using the model of the present invention and other models is shown, and is measured by the AUC index. The larger the AUC, the better the fitting effect. Figure 2 It can be seen that the polygenic risk score PRS obtained by the model of the present invention has a better prediction effect on T2D than the LDpred-inf method using only a single phenotype, and is also better than the SMTpred method. The SMTpred method is another new method in the world that improves the prediction ability of the target phenotype by integrating GWAS data of multiple phenotypes, but this method does not take advantage of cross-population data to improve the prediction ability of PRS. It can be seen that the method of the present invention improves the accuracy of specific PRS by integrating multi-population and multi-phenotype data, thereby improving the prediction accuracy of complex phenotypes or disease risks.

[0090] It should be noted that the polygenic risk scoring model based on the integration of multi-population and multi-phenotype data of the present invention can be applied to three or more populations and three or more phenotypes. The corresponding model construction method is the same as the above embodiment, and the calculation formula is an extension of the above calculation formula.

[0091] Example 2: The present invention is based on the construction system of a polygenic risk scoring model integrating multi-phenotype data of three or more populations.

[0092] It should be noted that among the three or more populations, one is the population to which the sample to be tested belongs, and the other populations are auxiliary populations. The overall parameter estimation method remains unchanged, and only the corresponding parameter matrix in step S3-4 in Example 1 needs to be expanded so that the parameter matrix includes the estimated values ​​of the heritability, genetic covariance, and environmental covariance of the population to which the sample to be tested belongs and other populations.

[0093] Specifically, it involves For heritability and genetic covariance, use subscripts 1 and 2 to represent the two phenotypes of the population to which the sample to be tested belongs, and use Denote the two phenotypes of the second population by Representing the two phenotypes of the kth population, we have: in, is the inverse of the covariance matrix of the site random effect; is the inverse of the environmental covariance matrix; X is the genotype matrix; It is The heritability of a phenotype; It is The phenotype and The genetic covariance of the phenotypes; is the number of individuals overlapping in the two studies; It is The phenotype and Environmental covariance of phenotypes; yes dimensional identity matrix. In addition, each parameter is labeled using a general expression, such as yes A special form of The phenotype and The genetic covariance of the phenotype, and the same goes for the rest.

[0094] Example 3: The present invention is based on the Edu polygenic risk scoring model construction system integrating multi-population and multi-phenotype data.

[0095] It should be noted that the method for constructing a polygenic risk score model integrating multi-phenotype data in the present invention can also be applied to any other phenotype besides T2D. For example, to calculate the polygenic risk score of adult education level (Edu), the North American population can be used as the target population, the European or Asian population can be used as the auxiliary population, and the Edu phenotype GWAS data can be collected. The auxiliary phenotypes can be selected from family income, occupation, height, etc., and modeling calculations are performed. The method is referred to Example 1 and Example 2.

[0096] Example 4: The present invention is based on the construction system of a polygenic risk scoring model integrating multi-population and multi-phenotype data.

[0097] The present invention also provides a PRS model construction system based on the integration of multi-population and multi-phenotype data, which provides the implementation environment required for the calculation of the method of the present invention, including a data entry and storage module, a data preprocessing module, a parameter estimation module, and a PRS calculation module, which are specifically introduced as follows.

[0098] The data entry and storage module is used to enter and store the phenotype GWAS data required for analysis of the ethnic group to which the sample to be tested belongs and other ethnic groups, as well as the reference genome data, according to a list of phenotypes to be tested consisting of the phenotypes to be tested and other phenotypes genetically correlated therewith, and to enter and store the genotype data of the sample to be tested.

[0099] The data preprocessing module is used to calculate the LD scores between two loci and the comprehensive LD score of each locus according to the reference genome data.

[0100] The parameter estimation module is used to estimate the parameters of the polygenic risk score model based on the comprehensive LD score and the phenotype GWAS data, including the heritability of each phenotype of the ethnic population to which the sample to be tested belongs and other ethnic populations, the genetic covariance and environmental covariance of the phenotype to be tested and each other phenotype, and the genetic covariance of each phenotype across populations, and estimate the site weight value of the polygenic risk score PRS.

[0101] The PRS calculation module is used to calculate the polygenic risk score PRS value of the sample to be tested based on the genotype data and the site weight value.

[0102] During specific implementation, the working process of the system of the present invention is as follows: A. Import and store data through the data entry storage module, including the following: A1. Phenotypic GWAS data: Collect the phenotype GWAS data required for analysis from the ethnic group to which the sample to be tested belongs and other ethnic groups, and import them into the database for storage; A2. Reference genome data: Collect reference genome data of the ethnic group to which the sample to be tested belongs and other ethnic groups, and store them in the database; A3. Genotype data of the sample to be tested: import external data, or obtain chip typing data through gene chip technology or high-throughput sequencing technology, import the chip typing data output by the gene detection equipment into the data entry module, and use Minimac4 to fill in the genes, and together with the chip typing data, form the genome data, and enter it into the database for storage.

[0103] B. Preprocess the data entered into the database through the data preprocessing module, including the following contents: B1. Format the phenotypic GWAS data required for analysis in the ethnic group to which the sample belongs and other ethnic groups. For GWAS data that do not provide a z-score, calculate the z-score based on its genetic effect value and the standard deviation of the genetic effect value; B2. Based on the reference genome data, calculate the LD score between every two sites and the comprehensive LD score of each site.

[0104] C. Calculate the parameters required for the model through the parameter estimation module, including the following: C1. Calculate the heritability of each phenotype of the ethnic group to which the sample belongs and other ethnic groups based on the comprehensive LD score and phenotype GWAS data; C2, based on the comprehensive LD score and phenotypic GWAS data, calculate the genetic covariance and environmental covariance between every two phenotypes; C3. Calculate the genetic covariance across populations for each phenotype based on the comprehensive LD score and phenotype GWAS data; C4. Calculate the site weight value of the polygenic risk score PRS based on the heritability of each phenotype of the ethnic population to which the sample to be tested belongs and other ethnic populations, the genetic covariance and environmental covariance between every two phenotypes, and the genetic covariance of each phenotype across populations.

[0105] D. Calculate the polygenic risk score PRS value of the sample to be tested through the PRS calculation module, including the following: D1, based on the PRS site weight value output by the parameter estimation module, calculate the polygenic risk score PRS value of the sample to be tested; D2. Output the polygenic risk score PRS value and the relevant information of the sample to be tested.

[0106] Furthermore, the system of the present invention can be installed on a server with strong computing power, which can be a local server or a cloud server.

[0107] Although the present invention has been described in specific embodiments, these embodiments are only used to explain the present invention and do not limit the invention. In any one or more embodiments or examples, the specific features, structures, materials or characteristics described may be combined in a suitable manner. After reading this specification, those skilled in the art may modify, replace and modify the embodiments as needed without violating the principles and purpose of the present invention. Although these changes do not make creative contributions, they are still protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data, characterized in that: The steps include: S1. According to the list of phenotypes to be tested consisting of the phenotypes to be tested and other phenotypes with genetic correlation with them, collect the phenotype GWAS data required for analysis in the ethnic population to which the sample to be tested belongs and other ethnic populations, as well as the reference genome data, and collect the genotype data of the sample to be tested; S2, based on the reference genome data collected in S1, calculate the LD score between every two sites and the comprehensive LD score of each site; S3. Based on the comprehensive LD score obtained in S2 and the phenotypic GWAS data collected in S1, estimate the parameters of the polygenic risk score model, including the heritability of each phenotype in the ethnic population to which the sample to be tested belongs and other ethnic populations, the genetic covariance and environmental covariance between every two phenotypes, and the genetic covariance of each phenotype across populations, and estimate the site weight value of the polygenic risk score PRS; S4. Calculate the polygenic risk score PRS value of the sample to be tested based on the genotype data collected in S1 and the site weight value obtained in S3.

2. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 1, characterized in that: The calculation formula of the LD score in S2 is as follows: in, , Respectively represent The site, individual sites; is the LD score of the two loci; , It is the first of two sites The genotypes of the samples; , is the average of the genotypes at the two loci.

3. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 2, characterized in that: The calculation formula of the comprehensive LD score in S2 is as follows: in, It is The comprehensive LD score of each locus; , They are The site, individual sites; is the LD score between the two loci.

4. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 3, characterized in that: The calculation formula for the heritability of each phenotype in S3 is as follows: in, It is The heritability of a phenotype; is the number of sites; It is Sample size for each phenotype; and, It is an estimate of the regression coefficient obtained by regressing the square z-score of the site in the GWAS data of the same population onto the LD score of the site. The calculation formula is as follows: in, It is The phenotype of GWAS z-score of each locus; It is The comprehensive LD score of each locus; It is an estimate of the intercept obtained by regressing the squared z-score of a site in GWAS data within the same population onto the LD score of the site.

5. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 4, characterized in that: The calculation formula of the genetic covariance between each two phenotypes in S3 is as follows: in, is the estimate of the genetic covariance between the first and second phenotypes; is the sample size of the first phenotype; is the sample size of the second phenotype; is the number of sites; and, It is the estimated value of the regression coefficient obtained by regressing the product of the z-score of each site in the GWAS data of the first phenotype in the ethnic population to which the sample to be tested belongs and the first phenotype in other ethnic populations onto the site LD score. The calculation formula is as follows: in, , is the first order of magnitude of the same phenotype in the ethnic group to which the sample belongs and in other ethnic groups. GWAS z-score of each locus; It is The comprehensive LD score of each locus; are the estimated values ​​of the regression coefficient and intercept, respectively.

6. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 5, characterized in that: The calculation formula of the environmental covariance between each two phenotypes in S3 is as follows: in, is the estimate of the environmental covariance between the first and second phenotypes; It is an estimate of the intercept obtained by regressing the z-score of the site in the GWAS data within the same population onto the LD score of the site; is the sample size of the first phenotype; is the sample size of the second phenotype; is the number of repeated individuals in the two phenotype GWAS data; is an estimate of the genetic covariance between the first and second phenotypes.

7. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 4, characterized in that: The calculation formula for the genetic covariance of each phenotype across populations in S3 is as follows: in, Is the first among the ethnic group of the sample to be tested and other ethnic groups Estimates of the cross-population genetic covariance of phenotypes; is the number of sites; It is The sample size of each phenotype in the ethnic group to which the sample to be tested belongs; It is The sample size of each phenotype in other ethnic groups; and, The first phenotypes and the first The estimated value of the regression coefficient obtained by regressing the product of the z-score of each site in the GWAS data of the phenotype to the site LD score is calculated as follows: in, , Is the first among the ethnic group of the sample to be tested and other ethnic groups The respective GWAS z-score of each locus; It is The comprehensive LD score of each locus; The first phenotypes and the first The intercept is estimated by regressing the product of the z-score of each locus in the GWAS data for the phenotype onto the locus LD score.

8. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 7, characterized in that: The calculation formula of the site weight of PRS in S3 is as follows: in, It is the first Phenotype The posterior mean of the estimated genetic effect values ​​for each locus; It is the first Phenotype The posterior mean of the estimated genetic effect values ​​for each locus; It is the first Phenotype The specific genetic effect value of each locus; It is the first Phenotype The specific genetic effect value of each locus; It is Phenotype The genotype standard deviation of each locus; is the number of sites; and, and The calculation method is as follows: in, is the locus-specific genetic effect value; is the genotype matrix of loci with population- and phenotype-specific genetic effects; is the phenotypic value after removing the influence of covariates; yes The inverse of the marginal distribution covariance matrix; is the posterior mean vector of the estimated random genetic effect values ​​of the loci; X is the genotype matrix; the symbol represents the Kronecker product; is the number of sites, yes dimensional identity matrix; is the inverse of the covariance matrix of the site random effect; is the inverse matrix of the environmental covariance matrix; and, , and The calculation method is as follows: in, yes Marginal distribution covariance matrix; yes The inverse of the marginal distribution covariance matrix; is the inverse of the covariance matrix of the site random effect; is the inverse of the environmental covariance matrix; X is the genotype matrix; the symbol represents the Kronecker product; is the number of sites, yes dimensional identity matrix, , , The subscripts correspond to the identity matrix of the dimension respectively; It is The heritability of a phenotype; It is Phenotype and The genetic covariance of each phenotype; is the number of individuals overlapping in the two studies; It is Phenotype and Environmental covariance of phenotypes; and, Approximately: Among them, the symbol and denote element-wise product and Kronecker product respectively; and is the LD matrix of all sites; and It is the LD matrix between all loci and large effect loci in the ethnic group to which the sample to be tested belongs and other ethnic groups; matrix and matrix are the standardized genotype matrices from the reference genomes of the ethnic group to which the sample to be tested belongs and other ethnic groups, with a mean of 0 and a variance of ; z is the z-score of the GWAS genetic effect value; are the GWAS sample sizes of phenotype 1 and phenotype 2 in the target population and the population to be tested, respectively; is the number of overlapping individuals in two GWAS studies of the same population.

9. The method for constructing a polygenic risk scoring model based on the integration of multi-population and multi-phenotype data according to claim 8, characterized in that: The calculation formula of the polygenic risk score PRS value in S4 is as follows: in, For new individuals or samples, PRS value under each phenotype; It represents the genotype data of a new individual, which contains the codes of multiple SNPs; is a vector, which refers to the estimates of the random genetic effects of all loci for a phenotype; It corresponds to the first A genotype data matrix for specific effect sites of each phenotype; is a vector containing Estimates of specific effect sizes associated with subsets of SNPs.

10. A polygenic risk scoring model construction system based on the integration of multi-population and multi-phenotype data, characterized in that: It includes data entry and storage module, data preprocessing module, parameter estimation module and PRS calculation module; The data entry and storage module is used to collect phenotype GWAS data required for analysis from the ethnic group to which the sample to be tested belongs and other ethnic groups, as well as reference genome data, and collect genotype data of the sample to be tested, based on a list of phenotypes to be tested consisting of the phenotypes to be tested and other phenotypes genetically correlated therewith; The data preprocessing module calculates the LD score between two loci and the comprehensive LD score of each locus according to the collected reference genome data; The parameter estimation module is used to estimate the parameters of the polygenic risk score model based on the obtained comprehensive LD score and the collected phenotype GWAS data, including the heritability of each phenotype of the ethnic population to which the sample to be tested belongs and other ethnic populations, the genetic covariance and environmental covariance between every two phenotypes, and the genetic covariance of each phenotype across populations, and estimate the site weight value of the polygenic risk score PRS; The PRS calculation module is used to calculate the polygenic risk score PRS value of the sample to be tested based on the collected genotype data and the obtained site weight value.

Images