Predictive Modeling of Microbial Community Data Using Phylogeny-Regularized Regression Models

• Activity Number: 225 - The Human Microbiome: From Discovery Studies to Statistical Predictive Personalized Medicine
• Type: Topic Contributed
• Date/Time: Monday, July 29, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Jun Chen*
• Companies: Mayo Clinic
• Keywords: microbiome; phylogeny; high-dimensional statistics; generalized linear mixed model; sparse regression model; predictive modeling

Abstract

Human microbiome studies have revealed an essential role of the human microbiome in health and disease, opening up the possibility of building microbiome-based predictive models. One unique characteristic of microbiome data is the phylogenetic tree that relates all the microbial taxa. It has frequently been observed that a cluster or clusters of taxa are associated with an outcome due to shared biological functions (clustered signal). Depending on the specific condition, a large or a small number of taxa can be involved, representing two distinct biological models (dense and sparse model). We thus develop “glmmTree”, a phylogeny-regularized generalized linear mixed model, for clustered and dense signal, and “SICS”, a phylogeny-regularized sparse generalized linear model, for clustered and sparse signal, respectively. glmmTree uses the global phylogeny-based similarity between microbiomes to predict the outcome while SICS performs variable selection and uses a novel phylogeny-based smoothness penalty to smooth the coefficients of related microbial taxa. Simulation studies and real data applications were used to demonstrate the performance of the proposed method.