Abstract

Bacterial dysbiosis has been implicated in various clinical conditions, e.g., caries, gut diseases and cancer. Microbial composition and abundance can be captured by microbiome DNA sequencing for the questions “what bacteria are there?”, while metatranscriptomics through RNA sequencing identifies functional characterization of complex microbial communities to answer, “what the bacteria do there”. The joint analyses of paired metagenomics and metatranscriptomics data, as well as with the metabolome data is one way to study bacterial species and genes’ functional roles in statuses of health and diseases, but remains challenging due to the high dimension and sparsity of the data. To address this knowledge gap, we developed a few statistical methods, including 1) a two-step approach "IntegRatio" to study the differential transcriptional activity by investigating the RNA/DNA ratios at species, 2) a BZINB-based network analysis to handle excess zeros in the data, and 3) an improved elastic net model (ENVIM) to use microbiome data to predict metabolites. These methods are applied in studying Early Childhood Caries (ECC) and Inflammatory bowl diseases (IBD).