Search algorithms for promotion of novel biomedical research

Abstract

A major problem plaguing the field of biomedical research is the tendency to perform safe, incremental research.This implies that biomedical researchers tend to collaborate with other researchers who are already "close" to them in their co-authorship networks or citation networks. We propose a novel method to find new potential collaborators in the Synergy research project. The idea is to include information from molecular networks and propose researchers working on "nearby" molecules as potential collaborators. Based on this idea, we have built a software application where a biomedical researcher could input a list of molecules of her choice and find several ranked lists of potential collaborators as output. The underlying network is the Synergy network - a multilayer network formed from data in the PubMed database. This Master's thesis describes the algorithms to find and sort these potential collaborators. These have been implemented in Java in the Synergy software application. The algorithms output the results within minutes even with tens of millions of author nodes and publication nodes in the network. Several potential collaborators are identified with sample lists of molecules in this thesis. Finally, this thesis validates the results found by comparing the probability of molecules which serve as research topics of two co-authors as being neighbors to the probability of molecules chosen at random being neighbors. It is found that the former is considerably higher than the latter.