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

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Transcriptome-wide association studies (TWASs) integrate GWAS and eQTL data to identify genes linked to complex traits.
• Existing TWAS methods often overlook gene-gene correlations and can produce false positives due to genetic variant effects.

Purpose of the Study:

• To introduce TWASKnockoff, a knockoff-based framework for robust causal gene-tissue pair detection.
• To address limitations of marginal association testing in current TWAS methods.

Main Methods:

• TWASKnockoff employs a knockoff inference approach to assess conditional independence between gene-trait pairs.
• It accounts for correlations in cis-predicted gene expression and between gene expression and genetic variants.
• A theoretical correlation matrix is estimated via parametric bootstrap, followed by knockoff-based inference to control the false discovery rate (FDR).

Main Results:

• TWASKnockoff demonstrates superior FDR control compared to traditional TWAS methods.
• The framework significantly improves power for detecting causal gene-trait associations at a fixed FDR level.
• Application to type 2 diabetes data validated its effectiveness.

Conclusions:

• TWASKnockoff provides a more accurate and powerful approach for identifying causal genes in complex trait association studies.
• The method offers improved control over false discoveries, enhancing the reliability of gene-trait pair nominations.