Testing Presence-Absence Association in the Microbiome Using LDM and PERMANOVA (306545)Yijuan Hu, Emory University
*Andrea Nicole Lane, Emory University
Glen A. Satten, Centers for Disease Control and Prevention
Keywords: microbiome, presence-absence
The difference between two groups of microbial composition profiles can be characterized by the presence-absence status of certain microbes. But there is a lack of methods that provide a global test of the presence-absence difference and tests of individual operational taxonomic units (OTUs) while accounting for biases induced by variation in sampling depth (i.e. library size). PERMANOVA is a commonly used distance-based method for testing the global hypothesis of any microbiome effect. The linear decomposition model (LDM) includes the global test and tests of individual OTU effects. We propose to rarefy the OTU table so that all samples have the same depth and then apply the LDM to the rarefied table or PERMANOVA to a presence-absence distance based on the rarefied table. We repeat the process for a number of randomly rarefied tables and sum up the test statistics over rarefactions. Our simulations indicate that the proposed strategy is robust to all systematic differences in library size. We also explored the optimal number of rarefactions that balance statistical power and computational cost and provide practical guidelines on how to select the rarefaction depth.