Leveraging the Discrete and Continuous Structure of Phylogenetic Trees for the Analysis of Metagenomic Data

• Activity Number: 117
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Frederick Matsen*+ and Steven N. Evans and Brian Claywell
• Companies: Fred Hutchinson Cancer Research Center and University of California, Berkeley and Fred Hutchinson Cancer Research Center
• Keywords: metagenomics ; phylogenetics ; ordination ; principal components analysis

Abstract

Evolution gives hidden structure to the genetic data used in studies of microbial communities sampled in bulk from an environment. It is therefore common and useful to take evolutionary history into account when analyzing such data. The underlying structure can be formalized as a phylogenetic tree, which has aspects that are both discrete (tree topology) and continuous (evolutionary time). In this talk I will describe a variant of principal components analysis that compares microbial abundances while taking both of these aspects of genetic relatedness into account. The successive principal components from this variant correspond to signed weighting functions on the tree; these functions indicate the degree to which various parts of the tree are associated with the heterogeneity seen between metagenomic samples. We show that this procedure has attractive theoretical properties and produces informative results when applied to data from human-associated microbial communities.