Variable Selection--Based Weighting Schemes for Rare Variants in Sequence Data

• Activity Number: 468
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Andrea Byrnes*+ and Mingyao Li and Michael C Wu and Fred Wright and Li Yun
• Companies: The University of North Carolina at Chapel Hill and University of Pennsylvania School of Medicine and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
• Keywords: statistical genetics ; rare variants ; next generation sequencing ; variable selection ; high dimensional data

Abstract

Recent advances in sequencing technologies allow us to assess rare genetic variants (RVs), i.e. variants with minor allele frequencies < 0.05. Limited power to detect association between individual RVs and human traits has motivated burden tests wherein the trait is regressed on the weighted average of multiple RVs across a region. Several such schemes have been proposed, but there is little consensus on which is the best. Here, we compare several weighting schemes using extensive simulations that mimic large sequencing studies of a quantitative trait (phenotype) under a variety of situations including single and multiple RVs that are associated with the trait in the same and opposite directions. We investigate existing phenotype-independent and -dependent schemes and apply variable selection to increase the number of zero weights. Our simulations show that phenotype-dependent weighting schemes have increased power over phenotype-independent schemes in most cases. We also show that variable selection can further increase the power to detect the effect of multiple RVs in phenotype-dependent schemes. Results on a real data set of ~2000 sequenced individuals will also be shown.