Function inference for interacting proteins through protein signatures

Abstract

Abstract<br><br>Signatures are functional sub-units of proteins’ primary structures. They include domains, motifs, and binding sites. Signatures play an important role in determining the interactions web of a particular protein and thereby, general or specific function(s) of the protein can be inferred. Commonality of signatures among proteins was used as the key information to predict the web of interactions of all proteins in S. cerevisiae, C. elegans, and H. sapiens and then to infer function for interacting proteins. Known signatures in the three studied organisms were retrieved from Pfam database. Next, each protein was represented by a feature vector, called signature profile, indicating the presence or absence of known signatures in the amino acid sequence of that protein. Then, the similarity between signature profiles was calculated based on the Binary Similarity Function (BSF) on a pair-wise fashion. The BSF, scored the protein pairs and if the score was higher than a threshold the two proteins in the pair were predicted to be interacting. The threshold was specified in association with a confidence level. In cases where one protein in the pair has known function, the interacting partner was assigned a similar function. As proteins interact with more than one protein in cellular systems, the partner with the highest BSF score was considered as the most likely, and the function was inferred accordingly. Based on this methodology, 1024 new interactions were identified in C. elegans genome, as an example, and the unknown proteins involved in these interactions were categorized into 25 general functions.