Analyzing Matched Sets of Microbiome Data Using LDM

• 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): Yijuan Hu* and Zhengyi Zhu and Caroline Mitchell and Glen Alan Satten
• Companies: Emory University and Emory University and Vincent Center for Reproductive Biology, Massachusetts General Hospital, Harvard Medical S and Centers for Disease Control and Prevention
• Keywords: association test; clustered data; differentially abundant OTUs; paired data; 16S rRNA data; tests of individual taxa

Abstract

Matched data arise frequently in microbiome studies. For example, we may collect paired microbiome samples pre and post treatment from a set of individuals or matched case and control samples from different individuals. We have previously developed the linear decomposition model (LDM) for independent samples that provides both global test of any effect of the microbiome on the trait of interest and tests of the effects of individual OTUs with false discovery rate(FDR)-based correction for multiple testing. Here we make important extensions of the LDM to analyzing matched-set data. The baseline microbiome characterizing a set is treated as a ``nuisance parameters", allowing all efforts to focus on the common differences within sets. The sample correlations are accounted for via restricted permutation within sets. Our method can allow for discrete and/or continuous variables to be tested and within-set covariates (confounders) to be adjusted, and can accommodate unbalanced data (i.e., uneven number of samples per set). The flexibility of our method for a variety of matched-set designs is illustrated through the analysis of data from two microbiome studies.