PROSPECT improves cis-acting regulatory element prediction by integrating expression profile data with consensus pattern searches

Consensus pattern and matrix-based searches designed to predict cis-acting transcriptional regulatory sequences have historically been subject to larg e numbers of false positives. We sought to decrease false positives by inco rporating expression profile data into a consensus pattern-based search met hod. We have systematically analyzed the expression phenotypes of over 6000 yeast genes, across 121 expression profile experiments, and correlated the m with the distribution of 14 known regulatory elements over sequences upst ream of the genes. Our method is based on a metric we term probabilistic el ement assessment (PEA), which is a ranking of potential sites based on sequ ence similarity in the upstream regions of genes with similar expression ph enotypes. For eight of the 14 known elements that we examined, our method h ad a much higher selectivity than a naive consensus pattern search. Based o n our analysis, we have developed a web-based tool called PROSPECT, which a llows consensus pattern-based searching of gene clusters obtained from micr oarray data.