Adjusting for Overdispersion in Microbiome Data Analysis

• Activity Number: 323 - Recent Methods and Tools in Analyzing Human Microbiome Data
• Type: Topic Contributed
• Date/Time: Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Ekaterina Smirnova* and Glen Satten
• Companies: University of Montana and Centers for Disease Control and Prevention
• Keywords: microbiome ; normalization ; overdispersion ; 16S ; count data

Abstract

Microbiome next generation sequencing experiments measure counts of DNA fragments for a large number of species in a sample. These data are often analyzed by first calculating a distance matrix between samples, followed by ordination (using principal components of the distance matrix to represent each sample in a low-dimensional space). However, systematic differences in factors like library size (the number of counts per sample) or the extent of overdispersion can make groups with the same species frequencies to appear to be different after ordination. Simple log and power data transformations do not solve the problem. We propose a data adjustment approach based on estimating the overdispersion parameter in Poisson and Dirichlet mixture models and then standardizing the data to remove the overdispersion.