An Integrative Testing Procedure in SNP-Sets Analysis

• Activity Number: 42 - Statistical Genetics I – New Approaches for Association Mapping
• Type: Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Yeonil Kim* and Judong Shen and Yueh-Yun Chi and Fei Zou
• Companies: Merck & Co., Inc and Merck & Co., Inc. and University of Florida and University of North Carolina at Chapel Hill
• Keywords: SNP-set; genetic model; efficient resampling; sum test

Abstract

Statistical tests in single SNP or SNP-set analyses are typically performed with an assumed mode of inheritance for each individual SNP. However, in practice, their underlying genetic models are often unknown. When the employed model is deviated from the underlying genetic model, power may be greatly reduced. To overcome the limitation, an integrative association test , which can estimate the underlying genetic model from GWAS data, has been proposed in single SNP analysis by Kim et al. (2019). We extend the test procedure to a SNP-set analysis where joint effects of multiple SNPs are simultaneously tested. To aggregate effects of multiple SNPs, we consider the sum of maximum of score statistics over a range of candidate genetic models under a generalized linear model framework. The null distribution of the aggregated test statistic is obtained by extending the efficient resampling method. Using simulated data based on the Internal HapMap project and compound symmetry, we find that the new method well maintains the Type I error rate and can have improved power over existing multi-marker tests. The new method are applied to the Drosophila melanogaster Genetic Reference Panel data.