EST databases as multi-conditional gene expression datasets

Large-scale expression data, such as that generated by hybridization to microarrays, is potentially a rich source of information on gene function and regulation. By clustering genes according to their expression profiles, groups of genes involved in the same pathways or sharing common regulatory mechanisms may be identified. Publicly-available EST collections are a largely unexplored source of expression data. We previously used a sample of rice ESTs to generate 'digital expression profiles' by counting the frequency of tags for different genes sequenced from different cDNA libraries. A simple statistical test was used to associate genes or cDNA libraries having similar expression profiles. Here we further validate this approach using larger samples of ESTs from the UniGene projects (clustered human, mouse and rat ESTs). Our results show that genes clustered on the basis of expression profile may represent genes implicated in similar pathways or coding for different subunits of multi-component enzyme complexes. In addition we suggest that comparison of clusters from different species, may be useful for confirmation or prediction of orthologs.