Statistical Models for High-Dimensional Compositional Data with Applicatins to Microbiome Data

• Activity Number: 658
• Type: Invited
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Hongzhe Li*+
• Companies: University of Pennsylvania
• Keywords: Constrained estimation ; Bacterial network ; Simplex data ; Variable selection ; Metagenomics ; Phylogenetics

Abstract

With the development of next generation sequencing technology, researchers have now been able to study the microbiome composition using direct sequencing, whose output are bacterial taxa counts for each microbiome sample. One goal of microbiome studies is to associate the microbiome composition with environmental covariates or clinical outcomes, including (1) identification of the biological/environemental factors that are associated with bacterial compositions; (2) identification of the bacterial taxa that are associated with clinical outcomes. Statistical models to address these problems need to account for the high-dimensional and sparse and compositional nature of the data. In addition, the prior phylogenetic tree among the bacterial species provides useful information on bacterial phylogeny. In this paper, I present models with parameter constraints for identifying the bacteria that are associated with clinical outcomes and for studying the bacterial networks. I demonstrate these methods with a human gut microbiome study.