Prediction of Co-Regulated Gene Groups through Gene Ontology

Gene ontology (GO) is organized in three principles, cellular component, biological process and molecular function. analysis of GO annotations of a list of differentially expressed genes on microarrays became a common approach in helping with their biological interpretation. Earlier studies in GO analysis are based on a single principle, mostly Biological Process; valuable information in the other two principles is neglected. This paper proposes a novel approach to investigate gene co-regulation based on GO annotations from all three principles. We used the semantic similarity of GO annotations as a measure to partition genes into functionally related clusters and developed a performance index (PI) that consolidates GO annotations from all three principles to measure the quality of each cluster. We successfully applied our algorithm to yeast dataset. Our results indicate that PI is a good measure of the likelihood of a cluster being co-regulated by one or more TFs. Another analysis based on individual GO principle indicates that gene annotations in biological process are the most informative and those in cellular component are the least informative with regard of gene co-regulation. However, none of the analyses based on an individual principle could provide satisfactory classification. It is important to consider gene annotations in all three principles