Proteins that interact physically tend to be involved in the same cellular functions. Therefore protein-protein interactions can be used for in silico functional characterization. In this work, the InWeb, an inferred human interactome, is applied to three distinct problems. (1) We use the resource to create a repository of disease-related protein complexes. By combining this repository with tissue-wide gene expression data, we generate tissue-specific expression profiles for the disease-related protein complexes. Hereby, we are able to show that tissue-specific overexpression of disease complexes correlate with pathologic manifestations. (2) We use the inferred interactome as an integrative framework for prioritization the entire genome allowing us to select strong candidate genes for complex psychiatric diseases. We combine the protein-protein interaction network with heterogeneous data from genome-wide association (GWA) studies, text-mined phenotype association, linkage studies, and gene expression studies. This results in a combined prioritization of the genome, and allows us to identify a shortlist of candidate genes. We validate one top-scoring candidate (YWHAH) in 2,000 individuals and show that indeed it is associated to bipolar disorder. (3) We combine the protein-protein interaction network with drug-target and pharmacogenetic interactions in a predictive pipeline that can used to prioritize genes in pharmacogenomics studies. The predicted rank can be used to adjust prior probabilities in pharmacogenomics GWA studies and thereby reduce the inherent noise. In conclusion, this work shows that protein-protein interactions constitute a mature data source in computational biology that can be combined with tissue-, disease- and drug-related data to answer diverse scientific problems.

---

Everybody is welcome. Registration is not necessary.

---