Abstract

From an RNAi screen for genes that repress apoptosis, we have identified a set of about 130 candidates (see abstract by McCloskey et al.). We are using computational methods in an effort to predict functional networks through which these genes may act and to identify additional candidates and their relationships within the postulated networks.

We are initially working with a number of datasets as the basis for our predictions: microarray expression data, the physical orientation of genes within the worm genome, phylogenetic profile data, and PSORT predictions of protein localization. The datasets are normalized according to their overlap with the worm interactome and Wormbase subcellular localization data. They are then combined using a naive Bayesian network, to yield a set of potential functionally interacting protein pairs along with a likelihood score for each pair. Finally, we construct network maps of the protein pairs that fall above a desired likelihood ratio. Our initial efforts suggest a cluster of functional interactions that are predicted from the gene set identified by the RNAi screen; refinements of the computational methods are in progress.