A systematic assessment of phylogenomic approaches for microbial species tree reconstruction

A key challenge in microbial phylogenomics is that microbial gene families are often affected by extensive horizontal gene transfer (HGT). As a result, most existing methods for microbial phylogenomics can only make use of a small subset of the gene families present in the microbial genomes under consideration, potentially biasing their results and affecting their accuracy. To address this challenge, several methods have recently been developed for inferring microbial species trees from genome-scale datasets of gene families affected by evolutionary events such as HGT, gene duplication, and gene loss. In this work, we use extensive simulated and real biological datasets to systematically assess the accuracies of four recently developed methods for microbial phylogenomics, SpeciesRax, ASTRAL-Pro 2, PhyloGTP, and AleRax, under a range of different conditions. Our analysis reveals important insights into the relative performance of these methods on datasets with different characteristics, identifies shared weaknesses when analyzing complex biological datasets, and demonstrates the importance of accounting for gene tree inference error/uncertainty for improved species tree reconstruction. Among other results, we find that (i) AleRax, the only method that explicitly accounts for gene tree inference error/uncertainty, shows the best species tree reconstruction accuracy among all tested methods, (ii) PhyloGTP (developed previously by the authors of this paper) shows the best overall accuracy among methods that do not account for gene tree error and uncertainty, (iii) ASTRAL-Pro 2 is less accurate than the other methods across nearly all tested conditions, and (iv) explicitly accounting for gene tree inference error/uncertainty can lead to substantial improvements in species tree reconstruction accuracy. Importantly, we also find that all methods, including AleRax and PhyloGTP, are susceptible to biases present in complex real biological datasets and can sometimes yield misleading phylogenies.