SANA: separating the search algorithm from the objective function in biological network alignment, Part 1: Search

TL;DR

This paper demonstrates that the simulated annealing algorithm SANA significantly outperforms other search algorithms in biological network alignment, providing a systematic approach to evaluate and improve objective functions independently.

Contribution

The paper introduces SANA as a superior search algorithm for network alignment, separating search from objective functions to enable clearer evaluation and development.

Findings

SANA outperforms existing search algorithms across various objectives.

Simulated annealing with SANA is highly effective for network alignment.

The approach facilitates independent assessment of objective functions.

Abstract

Biological network alignment is currently in a state of disarray, with more than two dozen network alignment tools having been introduced in the past decade, with no clear winner, and other new tools being published almost quarterly. Part of the problem is that almost every new tool proposes both a new objective function and a new search algorithm to optimize said objective. These two aspects of alignment are orthogonal, and confounding them makes it difficult to evaluate them separately. A more systematic approach is needed. To this end, we bring these two orthogonal issues into sharp focus in two companion papers. In Part 1 (this paper) we show that simulated annealing, as implemented by SANA, far outperforms all other existing search algorithms across a wide range of objectives. Part 2 (our companion paper) then uses SANA to compare over a dozen objectives in terms of the biology they recover, demonstrating that some objective functions recover useful biological information while others do not. We propose that further work should focus on improving objective functions, with SANA the obvious choice as the search algorithm.