Abstract

Human microbiome has been shown to be associated with many complex human diseases, and  microbiome research has become more data-intensive and technology-driven. In order to make larger scientific contributions, statisticians should learn enough domain knowledge, become familiar with various databases and computational skills. In addition, sampling and measurement processes should determine the statistical analyses. In this talk, I will present three recent projects in the area of microbiome data analysis, including quantifying bacterial growth dynamics using shotgun metagenomic data, compositional mediation analysis and balance-based regression. I will illustrate these methods using various microbiome studies at the University of Pennsylvania.