Bioinformatic Prediction of Yeast Gene Function

The bioinformatic prediction of gene function is, although young, already an extensive field, and with the high quality of the yeast genome sequence and the already large and rapidly growing volume of yeast functional genomics data, the prediction of yeast gene function is a substantial subfield in itself. A wide variety of approaches have been developed to predict gene function, ranging from sequence analyses to assign genes into functional families (Bork and Koonin, 1998; Ponting, 2001; Bateman et al., 2004), to structural analyses to assign protein folds (Honig, 1999; Schonbrun et al., 2002; Godzik, 2003) and active sites (Fetrow and Skolnick, 1998; Madabushi et al., 2002), to phylogenetic analyses for subdividing gene families into functional subgroups (Eisen, 1998a, b; Abhiman and Sonnhammer, 2005; Engelhardt et al., 2005) or predicting interacting partners (Pazos and Valencia, 2002). As ‘gene function’ takes such a wide variety of forms, from the corresponding protein’s biochemical activity to its physical interaction partners to membership in a given pathway, we focus here only on the latter ‘network’ aspects of gene function: a protein’s interaction and pathway partners, and the inferences of function that derive from these. B io in fo rm a ti c P re d ic ti o n o f