Association Testing and Feature Selection for Microbiome and Host Genomics

• Activity Number: 234 - Novel Statistical Methods for High-Dimensional Microbiome and Metagenomics Data Analysis
• Type: Topic Contributed
• Date/Time: Monday, July 29, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Anna Plantinga* and Michael C. Wu
• Companies: Williams College and Fred Hutchinson Cancer Research Center
• Keywords: microbiome; lasso; genomics; kernel RV

Abstract

The human microbiota play an important role in health and disease, but the mechanism of association is often not known. Recent studies suggest that host genetics may affect microbiome composition, which may also be associated with gene expression. We propose a sparse kernel RV approach that tests for overall association between the microbiota and genomic features, such as host genetic variants or gene expression, while simultaneously shrinking the contribution of some genomic features to zero. The method is assessed using both simulated and real microbiome and genomic data.