Conflict in ant phylogeny results from complex interaction between multiple evolutionary signals and tree reconstruction artifacts

Ants, Formicidae, are a group of small social insects that inhabit nearly all terrestrial environments. Three competing hypotheses of ant relationships have been proposed, differing in the placement of Martialinae, a subfamily of cryptic, endogean ants. We used BUSCO genes to investigate the signals in individual and concatenated gene datasets. We found that gene trees support all three hypotheses. After concatenation, the three signals persist but their relative strength is model dependent. The CAT-posterior mean site frequencies (CAT-PMSF) approach (which our model adequacy tests show best explains the across-site compositional heterogeneity of the data) finds Martialinae as the sister of all ants but Leptanillinae. We tested the effect of across-lineage compositional heterogeneity using data-recoding and excluding highly heterogeneous taxa. These tests did not lead to the emergence of significant support for alternative tree topologies. However, we identified strong gene- and site-discordance in the data, and evidence that signals representing incongruent evolutionary processes exist in ant genomes supporting all three hypotheses. Incomplete lineage sorting and/or introgression seem to have significantly affected early ant evolution, which might make it impossible to establish whether Leptanillinae, Leptanillinae plus Martialinae, or Martialinae represents the sister of all the other ants.Supplying Data:5,991 single copy orthology genes identified by BUSCO, across 101 taxa.Our final dataset, containing 1,442 single copy orthology gene alignments from 55 taxa.1,442 individual gene tree analyses, under the best fitting model including LG+F+G+CXX.Concatenated supermatrix of 1442 genes (101 and 55 taxa), and astral consensus trees of these two datasets.Concatenated phyogentic anaysis (144gene-55taxa, 10 jackknifes), under GTR+G/LG+G/LG+F+G+CXX/CAT-PMSF models.35 genes passed AU test, and their concatenated phylogenetic analysis under CAT-Poisson model.Three single-signal datasets, and their concatenated phylogenetic analyses under LG+G/LG+F+G+CXX/CAT-PMSF models.Taxa removal analyses: concatenated phyogeny (144gene-55taxa, 10 jackknifes, taxa-removal), under LG+G/LG+F+G+CXX/CAT-PMSF models.Recoded analyses: concatenated phyogeny (144gene-55taxa, 10 jackknifes, recoded), under GTR+G/CAT-Poisson models.