Systems biology A multiobjective memetic algorithm for PPI network alignment Connor Clark* and Jugal Kalita Department of Computer Science, University of Colorado Colorado Springs, Colorado Springs, CO 80918, USA *To whom correspondence should be addressed. Associate Editor: Igor Jurisica Received on July 28, 2014; revised on December 10, 2014; accepted on January 27, 2015 Abstract Motivation: There recently has been great interest in aligning protein–protein interaction (PPI) networks to identify potentially orthologous proteins between species. It is thought that the topo- logical information contained in these networks will yield better orthology predictions than sequence similarity alone. Recent work has found that existing aligners have difficulty making use of both topological and sequence similarity when aligning, with either one or the other being better matched. This can be at least partially attributed to the fact that existing aligners try to combine these two potentially conflicting objectives into a single objective. Results: We present Optnetalign, a multiobjective memetic algorithm for the problem of PPI net- work alignment that uses extremely efficient swap-based local search, mutation and crossover op- erations to create a population of alignments. This algorithm optimizes the conflicting goals of topological and sequence similarity using the concept of Pareto dominance, exploring the tradeoff between the two objectives as it runs. This allows us to produce many high-quality candidate align- ments in a single run. Our algorithm produces alignments that are much better compromises be- tween topological and biological match quality than previous work, while better characterizing the diversity of possible good alignments between two networks. Our aligner’s results have several interesting implications for future research on alignment evaluation, the design of network align- ment objectives and the interpretation of alignment results. Availability and Implementation: The Cþþ source code to our program, along with compilation and usage instructions, is available at https://github.com/crclark/optnetaligncpp/ Contact: connor.r.clark@gmail.com Supplementary information: Supplementary data are available at Bioinformatics online. 1 Introduction As the sizes of known protein–protein interaction (PPI) networks grow, so does interest in analyzing them. One of the more ambitious efforts in this area is aligning the PPI networks of two different spe- cies, with the goal of identifying orthologous proteins as well as shared pathways and complexes that hint at the PPI network of a common ancestor. There has already been a good deal of progress in this area, with past work reporting success in finding large shared subnetworks between Saccharomyces cerevisiae and Homo sapiens, as well as success in reconstructing phylogeny based on network overlap discovered by an aligner (Kuchaiev and Przulj, 2011; Kuchaiev et al., 2010). Despite these successes, PPI network alignment is a young re- search area. The incomplete, noisy nature of existing PPI networks makes alignment difficult, with existing aligners performing dramat- ically better on noise-free synthetic networks compared with noisy real-world networks (Clark and Kalita, 2014). Furthermore, it has been found that the two objectives of biological and topological fit that these aligners optimize conflict to a larger extent than previ- ously realized. Different aligners construct alignments that are dra- matically different, with some optimizing topological fit at the V C The Author 2015. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oup.com 1988 Bioinformatics, 31(12), 2015, 1988–1998 doi: 10.1093/bioinformatics/btv063 Advance Access Publication Date: 9 February 2015 Original Paper Downloaded from https://academic.oup.com/bioinformatics/article/31/12/1988/213937 by guest on 21 April 2021