RECOMP: A Parsimony-Based Method for Detecting Recombination

The central role phylogeny plays in biology and its pervasiveness in comparative genomics studies have led researchers to develop a plethora of methods for its accurate reconstruction. Most phylogeny reconstruction methods, though, assume a single tree underlying a given sequence alignment. While a good first approximation in many cases, a tree may not always model the evolutionary history of a set of organisms. When events such as interspecific recombi nation occur, different regions in the alignment may have different underlying trees. Accurate reconstruction of the evolutionary history of a set of sequences requires recombination detection, followed by separate analyses of the nonrecombining regions. Besides aiding accurate phylogenetic analyses, detecting recombination helps in understanding one of the main mechanisms of bacterial genome diversification. In this paper, we introduce RECOMP, an accurate and fast method for detecting recombination events in a sequence alignment. The method slides a fixed-width window across the alignment and determines the presence of recombination events based on a combination of topology and parsimony score differences in neighboring windows. On several synthetic and biological datasets, our method performs much faster than existing tools with accuracy comparable to the best available method.