B0741
Title: Bayesian variable selection for interval-censored outcomes in genetic association studies
Authors: Ryan Sun - University of Texas MD Anderson Cancer Center (United States) [presenting]
Jaihee Choi - Rice University (United States)
Abstract: In genetic association studies, many disease risk loci will harbour large numbers of individual genetic mutations, all demonstrating strong association with the disease. While a small number of these mutations possess functions that truly affect disease risk, many variants are also non-functional mutations that are simply correlated with the causal variant. It is important to distinguish between the two types of mutations to advance translational goals, such as designing new therapies or stratifying high-risk subjects. Bayesian variable selection procedures are popular tools for identifying functional mutations when the outcomes of interest are continuous or binary. However, less attention has been focused on interval-censored outcomes, even though many of the richest publicly available genetic datasets provide large amounts of interval-censored data. The proportional hazards formulation combines a conventional spike and slab prior with a nonparametric spline for the baseline hazard term. The procedure with an application to fractures in the UK Biobank is illustrated.
