Correlated sequence-signatures as markers of protein-protein interaction

As protein-protein interaction is intrinsic to most cellular processes, the ability to predict which proteins in the cell interact can aid significant ly in identifying the function of newly discovered proteins, and in underst anding the molecular networks they participate in. Here we demonstrate that characteristic pairs of sequence-signatures can be learned from a database of experimentally determined interacting proteins, where one protein conta ins the one sequence-signature and its interacting partner contains the oth er sequence-signature. The sequence-signatures that recur in concert in var ious pairs of interacting proteins are termed correlated sequence-signature s, and it is proposed that they can be used for predicting putative pairs o f interacting partners in the cell. We demonstrate the potential of this ap proach on a comprehensive database of experimentally determined pairs of in teracting proteins in the yeast Saccharomyces cerevisiae. The proteins in t his database have been characterized by their sequence-signatures, as defin ed by the InterPro classification. A statistical analysis performed on all possible combinations of sequence-signature pairs has identified those pair s that are over-represented in the database of yeast interacting proteins. It is demonstrated how the use of the correlated sequence-signatures as ide ntifiers of interacting proteins can reduce significantly the search space, and enable directed experimental interaction screens. (C) 2001 Academic Pr ess.