Gene order alignment on trees with multiOrthoAlign

Abstract

We relate the comparison of gene orders to an alignment problem. Our evolutionary model accounts for both rearrangement and content-modifying events. We present a heuristic based on dynamic programming for the inference of the median of three genomes and apply it in a phylogenetic framework. multiOrthoAlign is shown accurate on simulated and real datasets, and shown to significantly improve the running-time of DupLoCut, an "almost" exact algorithm based on linear programming, developed recently for the same problem.

Operon-based approach for the inference of rRNA and tRNA evolutionary histories in bacteria

“…We validate our new approach on simulated datasets (cherries and cherries with a neighbor), and we also test it on the same Bacillus dataset of 12 genomes used in [15,17]. Our results show that with our simulated data, BOPAL has the ability to infer events and ancestral gene orders with higher accuracy than other gene alignment algorithms.…”

“…We first use all-vs-all pairwise global alignments between the operons of the two genomes compared to identify orthologous operons. This is one of the major differences between our approach and the previous ones presented in [11,16,17]: instead of aligning the full gene orders to identify orthologous genes, we align only the operons, which tend to be more conserved. Moreover, the global alignments are not used to label events at this time, but only to find similar operons, which allows us to use a simpler scoring mechanism.…”

“…consecutive events on the same genes that do not directly appear on an alignment of the two genomes. This is another improvement over the previous algorithms, which were designed to consider only visible events [11,16,17]. If no match within the same genome is found with a score ≥ 0, we simply infer that the non-mapped operon was deleted in the other genome.…”

“…Andreotti et al [15] then proposed a faster and more efficient linear programming algorithm for the duplication-loss model, and generalized it to the median of three genomes setting. More recently, OrthoAlign [16] and multiOrthoAlign [17] were developed to generalize the evolutionary model to account for rearrangements (inversions and transpositions) in addition to duplications and losses. The idea there was to use dynamic programming to align the rRNA and tRNA gene orders and identify orthologs, and then explain the mismatches and gaps in the alignment by inferring rearrangement events (inversions and transpositions) and content-modifying events (duplications and losses).…”

“…Our results show that with our simulated data, BOPAL has the ability to infer events and ancestral gene orders with higher accuracy than other gene alignment algorithms. Similarly, on a biological dataset, BOPAL has performed equally as well as multiOrthoAlign [17] and DupLoCut [15] at generating the ancestral tree in a minimal amount of events.…”