Knockoff procedure improves susceptibility gene identifications in conditional transcriptome-wide association studies.

Abstract

Transcriptome-wide association studies (TWASs) have been developed to identify candidate genes associated with complex traits by integrating genome-wide association studies (GWASs) with expression quantitative trait loci (eQTL) data. However, most existing TWAS methods assess the marginal association between a single gene and a trait of interest, ignoring the influence of other genes in the same genomic region. Furthermore, false-positive gene-trait associations may arise due to correlations between eQTLs and nearby causal genetic variants. We introduce TWASKnockoff, a knockoff-based framework for detecting susceptibility genes using GWAS summary statistics and eQTL data. Unlike traditional TWAS approaches that rely on marginal testing, TWASKnockoff evaluates the conditional independence of each gene-trait pair, accounting for both cis-predicted expression correlations across genes and correlations between gene expression levels and genetic variants. TWASKnockoff estimates the correlation matrix of all genetic elements (including cis-predicted gene expression levels and genetic variant genotypes) by averaging estimations from parametric bootstrap samples, then applies knockoff-based inference to identify susceptibility genes while controlling the false discovery rate (FDR). Through simulations and an application to type 2 diabetes mellitus (T2D) data, we demonstrate that TWASKnockoff achieves superior FDR control and enhances power in detecting relevant gene-trait pairs at a fixed FDR level.