Analysis of Composition of Microbiomes (ANCOM): A novel method for studying microbial composition

Wednesday, September 24, 2014 — Poster Session IV
10:00 a.m. –12:00 p.m.	FAES Academic Center	NIEHS	MICROBIO-6

* FARE Award Winner

Authors

• S Mandal
• W van Treuren
• RA White
• M Eggesbo
• R Knight
• SD Peddada

Abstract

There is considerable interest in recent years to understand factors regulating human microbiota and the effect of the composition of microbiota on human health. Unlike the gene expression and other common high dimensional data, the microbiome data belongs to a simplex in the Euclidean space. Existing methods either discount the underlying compositional structure in the microbiome data (e.g. the t-test and ANOVA) or use inappropriate probability distributions (e.g. the multinomial and Dirichlet-Multinomial) which increase false discovery rate. Here we introduce a novel statistical framework called Analysis of Composition of Microbiomes (ANCOM) that accounts for the underlying compositional structure in the data. ANCOM makes no distributional assumptions, and is sufficiently general to enable easy adjustment for covariates. It is computationally fast to process thousands of taxa in a few minutes (e.g. 25 minutes to process 12000 OTUs on a Macbook Pro). Simulation studies reveal that ANCOM controls the FDR while the t-test and the recently published methodology called ZIG (Paulson et al., 2013) fail to do so with FDRs exceeding 75%. Our simulation studies also reveal that ANCOM is substantially more powerful than ZIG. We illustrate ANCOM using two publicly available microbial datasets in the human gut and soil.

back to top