Dimension Reduction of Longitudinal Microbiome Data via Tensor Functional SVD

• Activity Number: 432 - Novel Statistical Methods for Microbiome Data Analysis
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Pixu Shi* and Rungang Han and Anru Zhang
• Companies: Duke University and University of Wisconsin-Madison and Duke University
• Keywords: Microbiome; tensor; longitudinal ; dimension reduction

Abstract

The reduction of sequencing cost has prompted more microbiome studies with longitudinal measurements of bacterial abundance. Longitudinal microbiome data can often be formatted into a high-dimensional order-3 tensor with three modes representing the subject, time, and bacteria respectively. Since the time of measurement for different subjects can be highly variable, the values of such the order-3 tensor are typically not well-aligned, making it challenging to analyze the trajectory of bacterial abundance over time and identify key bacteria associated with time or clinical phenotypes. In this paper, we propose a new tensor functional SVD method that performs dimension reduction to assist the analysis of high-dimensional longitudinal microbiome data. The new method can extract the key components in the trajectories of bacterial abundance, identify representative bacterial taxa for these key trajectories, and group subjects based on the change of bacteria abundance over time. The new method is also flexible to handle microbiome measurements at irregular time points for different subjects.