Abstract The emergence of conflicting signals among different genomic regions, individual genes, and inference methods has been frequently reported in phylogenomic studies. Conflicts can occur from methodological error and biological processes, and disentangling the contribution of each putative source of discordance became a major challenge. Here, we assembled a dataset of 684 low‐copy nuclear markers from genome skimming data to build a phylogeny of Melastomataceae. The primary goal of this paper was to evaluate the phylogeny backbone regarding conflict across tribal relationships. To this end, we further investigated putative sources driving phylogenetic conflicts and uncertainty, including incomplete lineage sorting, nucleotide evolutionary rate heterogeneity, lack of information, inference error, and gene flow. The phylogenetic hypotheses of Melastomataceae estimated here (concatenated and coalescent) have corroborated the recurrent pattern found across several studies, in which major clades (subfamilies and tribes) have high support and a stable composition, but uncertainty is recovered for the relationships among some of them. The power of our explanatory variables for gene tree variation was effective to explain the variation in the backbone nodes. Inference error was the predictor explaining most of the variance found, a scenario that could be linked with an ancient rapid radiation in the family. The second most important predictor to explain gene tree variation was the ratio between the rates of molecular evolution of genes supporting a node versus conflicting signal. The signal of this predictor was particularly high in the node describing the sister relationship of two subfamilies (Olisbeoideae and Kibessioideae), indicating that long‐branch attraction artifacts might be affecting this deep relationship. Ancient gene flow was not estimated as a major factor driving backbone node uncertainty, but some signal of ancient introgression was recovered for specific nodes associated with a “hotspot” of conflicting relationships involving several tribes. Our analyses point to a scenario where lack of information does not seem to explain tribal relationship uncertainties with the data at hand. Therefore, attention should be given to better analyzing the molecular data available, especially with the emergence of new methodologies.
Reginato et al. (Mon,) studied this question.