A data integration approach to predict host-pathogen protein-protein interactions: application to recognize protein interactions between human and a malarial parasite.

Lack of large-scale efforts aimed at recognizing interactions between host and pathogens limits our understanding of many diseases. We present a simple and generally applicable bioinformatics approach for the analysis of possible interactions between the proteins of a parasite, Plasmodium falciparum, and human host. In the first step, the physically compatible interactions between the parasite and human proteins are recognized using homology detection. This dataset of putative in vitro interactions is combined with large-scale datasets of expression and sub-cellular localization. This integrated approach reduces drastically the number of false positives and hence can be used for generating testable hypotheses. We could recognize known interactions previously suggested in the literature. We also propose new predictions which involve interactions of some of the parasite proteins of yet unknown function. The method described is generally applicable to any host-pathogen pair and can thus be of general value to studies of host-pathogen protein-protein interactions.