A Nearly Optimal Sequential Testing Approach to Permutation-Based Association Testing

• Activity Number: 29 - SPEED: An Ensemble of Advances in Genomics and Genetics
• Type: Contributed
• Date/Time: Sunday, July 29, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Julian Hecker* and Ingo Ruczinski and Brent A. Coull and Christoph Lange
• Companies: Harvard T.H. Chan School of Public Health and Bloomberg School of Public Health and Harvard TH Chan School of Public Health and Harvard T.H. Chan School of Public Health
• Keywords: sequential testing; permutation; association testing; whole-genome sequencing; genetics

Abstract

While permutation-based approaches are ideal tools to address the statistical issues related to association testing of genetic variants in whole-genome sequencing (WGS) studies (e.g. potential violations of asymptotic distribution assumptions, non-normality of phenotypes and design imbalances), the actual application of such approaches is usually prohibitive due to the computational burden. Whereas current approaches use adaptive heuristics to reduce the number of permutations that are required in WGS studies, we propose a framework for permutation testing that is based on sequential testing theory and related to the Kiefer-Weiss problem. Our approach directly tests the permutation-based p-value against a pre-specified significance level. This procedure allows for the rigorous control of both error probabilities, type 1 and 2, and approaches the theoretical minimum of expected permutations. Our approach makes the application of permutation-based testing in WGS studies feasible and efficient in practice. In an application to a WGS study for a quantitative trait of lung function, we illustrate the performance of our approach and its practical implementation.