Estimating Regulatory Networks from Time Course Gene Expression Data via Adaptive Thresholding

• Activity Number: 129
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Sumanta Basu*+ and Ali Shojaie
• Companies: University of Michigan and University of Washington
• Keywords: Graphical model ; high dimensional learning ; Granger causality ; Vector autoregressive process ; gene regulatory network

Abstract

(Joint work with Ali Shojaie and George Michailidis)

Discovering regulatory interactions from time course gene expression data constitutes a canonical problem in functional genomics and systems biology. Time course gene expression data provide a dynamic view of expression levels of all the genes under study, and therefore, provides cues to the regulatory interactions among genes, which can be used to reconstruct the gene regulatory network. Using the framework of Granger causality, we propose a novel adaptive penalized estimator based on thresholded lasso estimates for estimation of regulatory interactions under different structural assumptions about the underlying time series. The proposed estimator offers significant improvement over available methods and have desirable asymptotic and small sample properties, which are explored through simulated data sets, as well as data from time course observations on T-cell activation.