Maximum likelihood phylogenetic reconstruction using gene order encodings

Gene order changes under rearrangement events such as inversions and transpositions have attracted increasing attention as a new type of data for phylogenetic analysis. Since these events are rare, they allow the reconstruction of evolutionary history far back in time. Many software have been developed for the inference of gene order phylogenies, including widely used maximum parsimony methods such as GRAPPA and MGR. However, these methods confronted great difficulties in dealing with emerging large nuclear genomes. In this study, we proposed three simple yet powerful maximum likelihood(ML) based methods for phylogenetic reconstruction by first encoding the gene orders into binary or multistate strings based on gene adjacency information presented in the given genomes and further converting these strings into molecular sequences. RAxML is at last used to compute the maximum likelihood phylogeny. We conducted extensive experiments using simulated datasets and found that although the multistate encoding is more complex and more time-consuming, it did not improve accuracy over the methods using simpler binary encodings. Among all methods tested in our experiments, MLBE is of the most accuracy in most cases and often returns phylogenies without errors. ML methods is also fast and in the most difficult case only takes up to three days to compute datasets with 40 genomes, making it very suitable for large scale analysis. We give three simple and robust phylogenetic reconstruction methods using different encodings based on maximum likelihood which has not been successfully applied for gene orderings before. Our development of these ML methods showed great potential in gene order analysis with respect to the high accuracy and stability, although formal mathematical and statistical analysis of these methods are much desired.