Simultaneous Optimization of Both Node and Edge Conservation in Network Alignment via WAVE

Abstract

Network alignment can be used to transfer functional knowledge between conserved regions of different networks. Typically, existing methods use a node cost function (NCF) to compute similarity between nodes in different networks and an alignment strategy (AS) to find high-scoring alignments with respect to the total NCF over all aligned nodes (or node conservation). But, they then evaluate quality of their alignments via some other measure that is different than the node conservation measure used to guide the alignment construction process. Typically, one measures the amount of conserved edges, but only after alignments are produced. Hence, a recent attempt aimed to directly maximize the amount of conserved edges while constructing alignments, which improved alignment accuracy. Here, we aim to directly maximize both node and edge conservation during alignment construction to further improve alignment accuracy. For this, we design a novel measure of edge conservation that (unlike existing measures that treat each conserved edge the same) weighs each conserved edge so that edges with highly NCF-similar end nodes are favored. As a result, we introduce a novel AS, Weighted Alignment VotEr (WAVE), which can optimize any measures of node and edge conservation, and which can be used with any NCF or combination of multiple NCFs. Using WAVE on top of established state-of-the-art NCFs leads to superior alignments compared to the existing methods that optimize only node conservation or only edge conservation or that treat each conserved edge the same. And while we evaluate WAVE in the computational biology domain, it is easily applicable in any domain.