# SANA NetGO: A combinatorial approach to using Gene Ontology (GO) terms to score network alignments

## TL;DR

This paper introduces NetGO, a new global measure for network alignment scoring that effectively weighs GO terms based on their frequency, improving correlation with alignment quality and taxonomic similarity.

## Contribution

NetGO is a novel, frequency-aware, global scoring measure for network alignments that outperforms existing GO-based methods in assessing alignment quality.

## Key findings

- NetGO correlates better with alignment quality than existing measures.
- NetGO effectively measures taxonomic similarity between species.
- Re-scoring of alignments shows NetGO's superior discrimination of good versus bad alignments.

## Abstract

Gene Ontology (GO) terms are frequently used to score alignments between protein-protein interaction (PPI) networks. Methods exist to measure the GO similarity between two proteins in isolation, but pairs of proteins in a network alignment are not isolated: each pairing is implicitly dependent upon every other pairing via the alignment itself. Current methods fail to take into account the frequency of GO terms across the networks, and attempt to account for common GO terms in an ad hoc fashion by imposing arbitrary rules on when to "allow" GO terms based on their location in the GO hierarchy, rather than using readily available frequency information in the PPI networks themselves. Here we develop a new measure, NetGO, that naturally weighs infrequent, informative GO terms more heavily than frequent, less informative GO terms, without requiring arbitrary cutoffs. In particular, NetGO down-weights the score of frequent GO terms according to their frequency in the networks being aligned. This is a global measure applicable only to alignments, independent of pairwise GO measures, in the same sense that the edge-based EC or S3 scores are global measures of topological similarity independent of pairwise topological similarities. We demonstrate the superiority of NetGO by creating alignments of predetermined quality based on homologous pairs of nodes and show that NetGO correlates with alignment quality much better than any existing GO-based alignment measures. We also demonstrate that NetGO provides a measure of taxonomic similarity between species, consistent with existing taxonomic measures--a feature not shared with existing GO-based network alignment measures. Finally, we re-score alignments produced by almost a dozen aligners from a previous study and show that NetGO does a better job than existing measures at separating good alignments from bad ones.