A comparison of three weighted human gene functional association networks

Abstract

Gene-gene association or protein-protein interaction databases have been important resource for the study of cellular functions and human diseases. A number of gene association databases have been available in the public domain. Each of these databases has its own unique virtues, but no single database could provide enough confidence and coverage. These years some meta-databases have been built by integrating various resources of gene functional associations and weighing the evidence of each association by some score systems. In this work, we compared three weighted genome-scale human gene association networks constructed from three such meta-databases, STRING, FunCoup and FLN, respectively. We found that the three networks share a large fraction of common genes but only quite limited overlapped interactions. However, most genes involved in important cellular processes and human diseases, as well as their pairwise interactions, is included in all of the three networks. This explains why all the three networks have been successfully applied in the study of cellular functions and diseases mechanisms. We believe that further integration of these meta-databases would provide higher confidence and coverage of gene associations in human proteome and facilitate the study of human gene association networks.