A method to adjust for the uncertainty of predicted expression in transcriptome-wide association study

Arunabha Majumdar,Tanushree Haldar

DOI: https://doi.org/10.1101/2022.11.06.515378

2024-06-20

Abstract:Transcriptome-wide association study (TWAS) is a promising approach to identifying novel genes associated with complex phenotypes. It offers better biological insights into the gene-phenotype associations than a genome-wide association study (GWAS). Standard TWAS approaches build a prediction model for the genetic component of expression based on reference transcriptome data. Next, an outcome phenotype is regressed on the predicted expression in separate GWAS data. The traditional unadjusted TWAS approach ignores the uncertainty of predicted expression, which can lead to unreliable discovery of gene-phenotype associations. We propose a novel approach to adjust for the uncertainty of predicted expression in TWAS. We adapt techniques from measurement error theory and implement bootstrapping algorithms for penalized regression to obtain an adjustment factor that needs to be incorporated in the unadjusted TWAS. We use extensive simulations to show that the unadjusted TWAS severely inflates the type 1 error rate, whereas the adjusted TWAS adequately controls it. At the expense of an alarmingly inflated false positive rate, the unadjusted TWAS offers a higher power than the adjusted TWAS, which is statistically unjustified. We demonstrate the merits of adjusted TWAS by conducting TWAS for height and lipid phenotypes while integrating the Geuvadis transcriptome and UK Biobank GWAS data.

Genomics