An alternative framework for transcriptome-wide association studies to detect and decipher gene-trait associations
Li, Z.; Gao, B.; Zhou, X.
Show abstract
Transcriptome-wide association studies (TWAS) are widely used to uncover transcriptomic mechanisms underlying disease. Here, we present VINTAGE, an alternative TWAS framework designed to identify and decipher gene-trait associations through two complementary tests: a genetic variance test that generalizes TWAS and unifies it with SKAT, offering a clearer understanding of TWAS false signals when expression is not relevant to SNP-trait associations; and a local genetic correlation test that distinguishes it from TWAS by explicitly quantifying and testing the proportion of genetic effects on trait mediated through gene expression. Applied to eQTLGen and eighteen traits from UK Biobank, VINTAGE improved gene-trait association power by an average of 5% and 87% over SKAT and TWAS, respectively. VINTAGE is also the only method effective in assessing potential gene mediation effects, revealing that most genes lack detectable mediation effects (median = 12%), which explains the power advantage of VINTAGE and SKAT over TWAS. Notably, VINTAGE identified 61 genes with significant mediation, highlighting the role of expression in genetic influences on traits.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Testing and controlling for horizontal pleiotropy with the probabilistic Mendelian randomization in transcriptome-wide association studies 98%
- OTTERS: A powerful TWAS framework leveraging summary-level reference data 97%
- Identification of putative causal loci in whole-genome sequencing data via knockoff statistics 97%
Similar papers in this journal
- Summary statistics from large-scale gene-environment interaction studies for re-analysis and meta-analysis 95%
- BEATRICE: Bayesian Fine-mapping from Summary Datausing Deep Variational Inference 95%
- CoMM-S2: a collaborative mixed model using summary statistics in transcriptome-wide association studies 95%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.