Reconstructing Ancestral Genomic Orders Using Binary Encoding and Probabilistic Models

Changes of gene ordering under rearrangements have been extensively used as a signal to reconstruct phylogenies and ancestral genomes. Inferring the gene order of an extinct species has the potential in revealing a more detailed evolutionary history of species descended from it. Current tools used in ancestral reconstruction may fall into parsimonious and probabilistic methods according to the criteria they follow. In this study, we propose a new probabilistic method called PMAG to infer the ancestral genomic orders by calculating the conditional probabilities of gene adjacencies using Bayes’ theorem. The method incorporates a transition model designed particularly for genomic rearrangement scenarios, a reroot procedure to relocate the root to the target ancestor that is inferred as well as a greedy algorithm to connect adjacencies with high conditional probabilities into valid gene orders. We conducted a series of simulation experiments to assess the performance of PMAG and compared it against previously existing probabilistic methods (InferCARsPro) and parsimonious methods (GRAPPA). As we learned from the results, PMAG can reconstruct more correct ancestral adjacencies and yet run several orders of magnitude faster than InferCARsPro and GRAPPA.