LAUSR

Phylogenetic inference based on genomic structural variations, that manipulate the gene order and content of whole chromosomes, promises to inform a more comprehensive understanding of evolution. The first challenge in using such data, the incompleteness of available de novo assemblies, is easing as long read technologies enable (near-)complete genome assembly, but methodological challenges remain. To obtain the input to rearrangement-based inference methods, we need to detect syntenic blocks of orthologous sequences, a task that can be accomplished in many ways, none of which are obviously preferable. In this paper, we use 94 reference quality genomes of primarily Mycobacterium tuberculosis (Mtb) isolates as a benchmark to evaluate these methods. The clonal nature of Mtb evolution, the manageable genome sizes, along with substantial levels of structural variation make this an ideal benchmarking dataset. We test several methods for detecting homology and obtaining syntenic blocks, and two methods for inferring phylogenies, comparing them to the standard method that uses substitutions for inferring the tree. We find that not only the choice of methods but also their parameters can impact results, especially among branches with lower support. In particular, a method based on an encoding of adjacencies applied to Cactus-defined blocks was fully compatible with the highly supported branches of the substitution-based tree. Thus, we were able to combine the two trees to obtain a supertree with high resolution utilizing both SNPs and rearrangements. Furthermore, we observed that the results were much less affected by the choice of the tree inference method than by the method used to determine the underlying syntenic blocks. Overall, our results indicate that accurate trees can be inferred using genome rearrangements, but the choice of the methods for inferring the homology matters and requires care.