Random Walks on Protein Interaction Networks

Milenkoski, Martin and Trivodaliev, Kire (2017) Random Walks on Protein Interaction Networks. In: PROCEEDINGS of the 14th Conference on Informatics and Information Technology. Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University in Skopje, Macedonia, Skopje, Macedonia, pp. 190-195. ISBN 978-608-4699-07-1


Download (342kB) | Preview
Official URL: http://ciit.finki.ukim.mk


Random walk is a stochastic process that describes a path which consists of a succession of random steps on some mathematical space. In terms of graphs, a random walk that starts from a given query node results in a stationary vector (distribution) which can be interpreted as the affinity of the query node being “connected” with other nodes in the graph. In this paper random walks are used for protein interaction networks analysis. More specifically, different types of random walks, such as simple random walk and random walk with restart are simulated on a weighted graph representation of a human protein interaction network. The performances of the variants of the random walk process are evaluated for topologically different query nodes in the graph, in terms of their convergence time and node coverage. Further comparison is done by analyzing the overlap in the stationary vectors of the random walks. An application of random walks in solving a biologically motivated problem is shown in predicting members of a partially known protein complex (complex membership problem). The input for this is a core set of proteins (i.e. the queries) making up a protein complex. The problem in terms of biology is whether this core set is complete or not. In this paper, potential members of protein complexes are found by ranking proteins according to their random walk based affinity to the query proteins in the partially known protein complex. Evaluation is done by using a leave-oneout method for experimentally confirmed protein complexes.

Item Type: Book Section
Subjects: International Conference on Informatics and Information Technologies > Students Session
Depositing User: Vangel Ajanovski
Date Deposited: 29 Nov 2017 18:27
Last Modified: 29 Nov 2017 18:27
URI: http://eprints.finki.ukim.mk/id/eprint/11402

Actions (login required)

View Item View Item