Phylogeny-Constrained Sparse Models for the Analysis of Microbiome 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): Jun Chen*+
• Keywords: Metagenomics ; Variable selection ; High-dimensional statistics

Abstract

Next generation sequencing technology opens a new era for microbiome research. One goal of microbiome studies is to identify microbial taxa that are associated with a biological or clinical outcome. Microbial taxa are related by a phylogenetic tree and closely related taxa tend to exhibit similar biological characteristics. Utilizing the phylogenetic tree information holds the key for more meaningful selection of microbial taxa. In this talk, I will present two phylogeny-constrained sparse models - sparse linear regression and sparse canonical correlation analysis that use the phylogenetic information. A phylogeny-constrained penalty function is used to impose certain smoothness on the linear coefficients according to the phylogenetic relationships among the taxa. Therefore, closely related taxa have a tendency to be selected together. Efficient coordinate descent algorithms are developed for optimization. Both simulations and real data applications show that phylogeny-constrained sparse models perform much better than the standard sparse models in identifying meaningful microbial taxa.