A gene-specific variance-control approach corrects polygenicity-driven inflation observed in transcriptome-wide association studies

We show that transcriptome-wide association studies (TWAS) and related methods (xWAS) using genetic predictors suffer from inflated false-positive rates for highly polygenic complex traits, increasing linearly with sample size and heritability. We propose a variance-control method that yields well-calibrated results when applied to both simulated and real data.