Predicting protein functions from redundancies in large-scale protein interaction networks

摘要: Interpreting data from large-scale protein interaction experiments has been a challenging task because of the widespread presence random false positives. Here, we present network-based statistical algorithm that overcomes this difficulty and allows us to derive functions unannotated proteins data. Our uses insight if two share significantly larger number common partners than random, they have close functional associations. Analysis publicly available Saccharomyces cerevisiae reveals >2,800 reliable associations, 29% which involve at least one protein. By further analyzing these tentative for 81 with high certainty. method is not overly sensitive positives in Even after adding 50% randomly generated interactions measured set, are able recover almost all (≈89%) original