Abstract

We present two new methods, xTADA and VBASS (Variational inference Bayesian ASSociation), that integrate expression data to improve power of rare variants association analysis. Optimized for bulk RNA-seq and single-cell transcriptomics data respectively, xTADA and VBASS model the association of disease risk as a function of expression profiles of relevant tissue or cell types in Bayesian frameworks. On simulated data, both methods show proper error rate control and better power than extTADA, the state-of-the-art Bayesian method. We applied the methods to published datasets and identified more candidate risk genes than extTADA with supports from literature or data from independent cohorts.