Statistical Modeling of Regulatory Networks with Applications in Integrative Genomics

Motivation: Over the course of a lifetime nearly 1 in 8 women will develop invasive breast cancer. Eighty-five percent of breast cancer cases are due to a lifetime of acquiring DNA mutations, which are known to foster tumor development by disrupting tightly regulated biological pathways. Aberrant amplifications and deletions of large stretches of DNA, known as copy number alterations (CNA), are one form of mutation ubiquitously manifested in breast cancer tumor cells. The focus of this work is two fold: (i) identifying which CNA dysregulate gene expression at the mRNA or protein level; (ii) also, uncovering which dysregulated biological pathways---represented by gene co-expression networks---exhibit the hallmarks of cancer progression.;Model: We developed a conditional graphical model, hereafter called spaceMap, for learning gene regulatory networks from multiple types of high-dimensional omic profiles. Through a penalized multivariate regression framework, spaceMap jointly models gene expressions as responses and CNA as predictors. In this setup, spaceMap infers an undirected network among gene expressions together with a directed network encoding which CNA dysregulate the gene-gene regulatory network. While the motivating application is learning cancer dysregulation, spaceMap can be applied generally to learn other types of regulatory relationships from high dimensional molecular profiles, especially those exhibiting hub structures.;Results: Simulation studies show spaceMap has greater power in detecting regulatory relationships over competing methods. Additionally, spaceMap includes a network analysis toolkit for biological interpretation of inferred networks. We applied spaceMap to the CNA, gene expression and proteomics data sets from CPTAC-TCGA breast (n=77) and ovarian (n=174) cancer studies. Each cancer exhibited disruption of 'ion transmembrane transport' and 'regulation from RNA polymerase II promoter' by CNA events unique to each cancer. Moreover, using protein levels as a response yields a more functionally enriched network than using RNA expressions in both cancer types. The network results also help to pinpoint crucial cancer genes and provide insights on the functional consequences of important CNA in breast and ovarian cancers.;Availability: The R package spacemap---including vignettes and documentation---is hosted at https://topherconley.github.io/spacemap.