Inferring Cis-region Hierarchies from Patterns in Time-Course Gene Expression Data

Abstract

Resolving the co-regulation relationships between genes is a major step toward understanding the underlying topology and dynamics of gene networks. Although co-expression of genes does not directly imply their co-regulation, model-based approaches coupled with the availability of large-scale gene expression data can help associate expression patterns with features in their cis-regions. Inspired by studies of transcriptional regulation in sea-urchin, here we report on preliminary validation of the following simple model for transcriptional regulation in yeast: the same Cis-Regulatory Modules (CRMs) in the cis-regions of different genes give rise to very similar functional events in the time-course expression profiles of those genes. We use a modified version of a prior algorithm for decomposing time-course gene expression patterns into functional events. To capture and reason about shared CRMs we introduce an order relationship, or a Regulation Hierarchy on the genes. When tested on actual time-course gene expression data of yeast preliminary results indicate 50% – 71% matches, of high confidence, between our derived and known cis-region regulation hierarchies. This hierarchy structure yields practical predictions when used with other type of genomic data, e.g. location of TF-DNA interactions.