DomainRank: Improving Biological Data Sets With Domain Knowledge and Google’s PageRank

Abstract

AbstractMotivationThe quality of biological data crucially affects progress in science. This quality can be improved with better measurement devices, more sophisticated experimental designs, or repetitious measurements. Each of these options is associated with substantial costs. We present a simple computational tool as an alternative. This algorithmic tool, called DomainRank, leverages simple domain knowledge and overlapping information content in biological network data to improve measurement quality at a negligible cost. Following the simple computational template of Domain-Rank, researchers can boost the confidence of their own data with little effort.ResultsWe demonstrate the performance of DomainRank in three test cases: DomainRank finds 14.9% more interactions in quantitative proteomics experiments, improves the precision of predicted residue-residue contacts from co-evolutionary data by up to 11.6% (averaged over 882 proteins), and identifies 89.2% more cross-links in photo-crosslinking/mass spectrometry (photo-CLMS) experiments. Although our proposed template is specialized on biological network data, we view this approach as an universal computational tool for data improvement that could be routinely applied in many disciplines.AvailabilityAn implementation of DomainRank is freely available: https://github.com/Rappsilber-Laboratory/pagerank-refine