Skip to Main Content
Recent high-throughput (HTP) technology has accumulated a large amount of protein-protein interaction (PPI) data. However, these data are laden with high false positive and false negative rate. Here we combine genomic data and topological metrics to eliminate false positive high-throughput PPIs, and utilize other genomic information that is not integrated in the previous stage to justify the outcome. We devise an approach to remove highly probable false positive HTP PPIs from HTP dataset. Our method takes the advantage of two acknowledged facts about PPI: (1) Two interacting proteins often show similar or related genomic features, for instance, similar gene expression profile; (2) The protein interaction network is believed to possess some certain topological properties. First, we calculate a confidence value for each HTP PPI according to gene co-expression, co-essentiality and times observed in different HTP experiments. Then, topological metrics is used to give a topological weight to an interaction. The two weights are summed up to represent initial weight of interactions in raw PPI network. Next, we use IRAP procedure to iteratively remove false positives. To demonstrate the usefulness of our approach, Gene Ontology-biological process annotation information is utilized to evaluate the ultimate refined PPIs.