Multivariable Association in Population-Scale Meta'omic Surveys

• 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): Himel Mallick* and Timothy Tickle and Lauren McIver and Gholamali Rahnavard and Long Nguyen and George Weingart and Siyuan Ma and Boyu Ren and Emma Schwager and Ayshwarya Subramanian and Joseph Paulson and Eric A. Franzosa and Hector Corrada Bravo and Curtis Huttenhower
• Companies: Merck & Co., Inc. and Broad Institute and Harvard University and Broad Institute and Massachusetts General Hospital and Harvard University and Harvard University and Harvard University and Harvard University and Broad Institute and Genentech and Harvard University and University of Maryland and Harvard University
• Keywords: Human Microbiome; Metagenomics; Microbiome Epidemiology; Differential Abundance Analysis; Multivariable Association Testing; Longitudinal Analysis

Abstract

It is challenging to associate features such as human health outcomes, diet, environmental conditions, or other metadata to microbial community measurements, due in part to their quantitative properties. Microbiome multi’omics are typically noisy, sparse (zero-inflated), high-dimensional, extremely non-normal, and often in the form of count or compositional measurements. Here we introduce an optimal combination of novel and established methodology to assess multivariable association of microbial community features with complex metadata in population-scale observational studies. Our approach, MaAsLin2, uses linear models to accommodate a wide variety of modern epidemiological study designs including cross-sectional and longitudinal. To construct this method, we conducted a large-scale evaluation of a broad range of scenarios under which straightforward identification of meta’omic associations can be challenging. These simulation studies reveal that MaAsLin2’s linear model preserves statistical power in the presence of repeated measures and multiple covariates while accounting for the nuances of meta’omic features and controlling false discovery.