LS-imputation: a new method to impute traits for genotyped individuals with GWAS summary data

Abstract

Genome-wide association study (GWAS) summary data have become extremely usefulin genetic research, largely facilitating the development of novel methods and new applications. However, a significant limitation of the current usage of GWAS summary data is its restriction to only linear single nucleotide polymorphism (SNP)-trait association analyses. To further expand the use of GWAS summary data, we propose the LS-imputation method, a nonparametric approach for large-scale imputation of the genetic component of a trait using only genotypic data and GWAS summary statistics for the trait of interest. With the imputed individual-level trait values and genotypes, it is possible to conduct any downstream analyses as if we have the individual-level GWAS data. Building upon this, we further extend the method to account for environmental effects by integrating GWAS summary statistics, individual-level genotypic data, and omic data for trait imputation. By leveraging SNP-trait and omics-trait association summary data, we can impute both the genetic and environmental components of a trait. This enhancement improves the accuracy of trait imputation, making it more suitable for analyses involving environmental variables. Lastly, to address the challenges of statistical inference posed by the correlation among imputed trait values, we propose a “divide and conquer” strategy that accounts for their covariance structure. We conduct both theoretical analyses and extensive real-data applications to demonstrate the effectiveness and advantages of the proposed methods.