Logistic Tree Gaussian Processes (LoTGaP) for the Microbiome

• Activity Number: 384 - Next-Generation Sequencing and High-Dimensional Data
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 12:00 PM to 1:50 PM
• Sponsor: Biometrics Section
• Author(s): Morris Greenberg* and Li Ma and Zhuoqun Wang and Pulong Ma and Anthony Sung
• Companies: Duke University and Duke University and Duke University and Duke University / SAMSI and Duke University Hospital
• Keywords: Gaussian processes; microbiome; Polya-Gamma augmentation; compositional data; Bayesian inference; phylogenetic trees

Abstract

Recently, there has been an increase in the number of studies tracking human microbiome compositions at irregular time points. We develop a new temporal model for compositional data to infer the dynamics of the microbiome in such settings, which we call logistic tree Gaussian processes (LoTGaP). Key features of LoTGaP include (1) using Gaussian processes to flexibly characterize the evolution of the microbiome composition over a finite set of days to handle missing/inconsistently measured data, (2) transforming microbial taxa abundance to log-odds on the internal nodes of the phylogenetic tree to address high-dimensionality, accelerate computation, preserve biological relationships, and allow for sparse data, and (3) allowing easy incorporation of fixed and random effects from covariates and design variables. We demonstrate our method through modeling and analyzing microbiome dynamics in a study of cancer patients undergoing hematopoietic cell transplantation.