Title: A statistical evaluation of whether animal models mimic human and a Bayesian hierarchical model for multi-omics integrative analysis
I will cover two recent projects in this talk. For the first topic, recent contradictory reports on whether mouse models mimic human in transcriptomic response (PNAS, 2013 3507-3512; PNAS, 2015 1167-1172) have created debates on usefulness of animal models. Instead of a dichotomous “yes”/“no” answer, we have developed a statistical evaluation framework with functional characterization for comparison of differential transcriptomic systems. I will present its application to: (1) various related inflammatory conditions (2) modENCODE life cycle data of C. elegant and D. melanogaster (3) TCGA breast cancer data. On the second topic, we consider multi-omics data integrative modeling by incorporating prior knowledge of a multi-layer overlapping group structure to improve variable selection in regression setting. In genomic applications, for instance, a biological pathway contains tens to hundreds of genes and a gene can contain multiple experimentally measured features (such as its mRNA expression, copy number variation and possibly methylation level of multiple sites). This biological prior knowledge contains hierarchical group structure as well as overlapping groups (e.g. two pathways can contain overlapped genes). We propose a Bayesian hierarchical indicator model that can conveniently incorporate the multi-layer overlapping group structure in variable selection. We discuss theoretical properties of the method. We apply the model to two simulations and one TCGA breast cancer example to demonstrate its superiority over other existing methods.