Waste Not, Want Not: Improved Normalization and Inference 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): Paul McMurdie*+
• Companies: Stanford University

Abstract

The development of culture-independent methods to assess microbiomes have revolutionized our understanding about the role of microbiomes inhabiting our planet as well as our own bodies -- creating a huge opportunity for clinical and ecological research. This data often arrives in the form of sample libraries that have been sequenced simultaneously and in parallel, but nevertheless vary widely in total number of sequences per sample. Unfortunately, the current practice for normalizing microbiome count data is inefficient in the statistical sense. Though not adopted in microbiome literature, well-established theory is available that simultaneously accounts for library size differences and biological variability using an appropriate mixture model. Moreover, specific implementations for DNA sequencing read count data (based on a Negative Binomial model for instance) are already available in RNA-Seq focused R packages such as edgeR and DESeq. Regarding microbiome sample-wise clustering, the most common procedure often discards samples that can be accurately clustered by alternative methods. I have provided microbiome-specific extensions for the R package, phyloseq.