De Novo Discovery of Cis-Regulatory Modules Using Time Course Gene Expression Data

Abstract

Time course microarray experiments are now a standard method for studying biological processes, a frequent aim of which is to discover regulatory relationships between genes. Regression-based methods have been proposed for de novo discovery of cis-regulatory sequences based on gene expression data. However, when applied to time course expression data, these methods model each individual time point separately, thereby ignoring the dynamic relationships between them. We explore an approach for joint modeling of promoter sequence data and time course expression data that capture such dynamic relationships. In higher order eukaryotic genomes most regulatory relationships are combinatorial. We propose a step-wise adaptive model building procedure that aims to capture such combinatorial cis-regulatory "modules". We apply our methods to the analysis of a microarray time-course experiment on Arabidopsis thaliana, some preliminary results show several known cis-regulatory elements as well as make some interesting predictions. This is joint work with Profs. Terry Speed and Mary Wildermuth at University of California, Berkeley.