Title: Visualization and unsupervised clustering of microbiome data
Authors: Yushu Shi - The University of Missouri Columbia (United States) [presenting]
Christine Peterson - UT MD Anderson Cancer Center (United States)
Liangliang Zhang - University of Texas MD Anderson Cancer Center (United States)
Robert Jenq - University of Texas MD Anderson Cancer Center (United States)
Kim-Anh Do - The University of Texas MD Anderson Cancer Center (United States)
Abstract: Microbiome plays an important role in human health and disease. We will first present aPCoA, an easy-to-use tool available as both an R package and a Shiny app, which improves data visualization by adjusting confounding covariates in a PCoA plot under non-Euclidean distance and enhances the presentation of the effects of interest. Then, we will briefly discuss commonly used metrics in unsupervised clustering of microbiome data and propose a new metric, which combines Bray Curtis and unweighted UniFrac distances and gives better performance on tested datasets. In the end, we will introduce a Bayesian model using Dirichlet tree multinomial mixture to cluster human microbiome data, which captures the tree-based topological structure of microbiome data and informatively selects tree nodes contributing to clustering.