Thomas Schlitt,1,3 Kimmo Palin,2 Johan Rung,1 Sabine Dietmann,1
Michael Lappe,1 Esko Ukkonen,2 andAlvis Brazma1
1European Bioinformatics Institute, EMBL-EBI, Wellcome Trust Genome Campus, Hinxton, Cambridge CB101SD, UK;
2Department of Computer Science, FIN-00014 University of Helsinki, Finland
We propose a novel method to identify functionally related genes based on comparisons of neighborhoods in gene networks. This method does not rely on gene sequence or protein structure homologies, and it can be applied to any organism and a wide variety of experimental data sets. The character of the predicted gene relationships depends on the underlying networks;they concern biological processes rather than the molecular function. We used the method to analyze gene networks derived from genome-wide chromatin munoprecipitation experiments, a large-scale gene deletion study, and from the genomic positions of consensus binding sites for transcription factors of the yeast Saccharomyces cerevisiae. We identified 816 functional relationships between 159 genes and show that these relationships correspond to protein–protein interactions, co-occurrence in the same protein complexes, and/or co-occurrence in abstracts of scientific articles. Our results suggest functions for seven previously uncharacterized yeast genes: KIN3 and YMR269W may be involved in biological processes related to cell growth and/or maintenance,
whereas IES6, YEL008W, YEL033W, YHL029C, YMR010W, and YMR031W-A are likely to have metabolic functions.
