Ultraconserved elements help resolve the phylogeny of an ancient radiation of venomous flies (Diptera: Asilidae)

Abstract

Robber flies or assassin flies (Diptera: Asilidae) are a diverse family of venomous predators. The most recent classification organizes Asilidae into 14 subfamilies based on a comprehensive morphological phylogeny, but many of these have not been supported in a subsequent molecular study using traditional molecular markers. To address questions of monophyly in Asilidae, we leveraged the recently developed Diptera-wide UCE baitset to compile seven datasets comprising 151 robber flies and 146 - 2,508 loci, varying in the extent of missing data. We also studied the behavior of different nodal support metrics, as the non-parametric bootstrap is known to perform poorly with large genomic datasets. Our ML phylogeny was fully resolved and well-supported, but partially incongruent with the coalescent phylogeny. Further examination of the datasets suggested the possibility that GC bias had influenced gene tree inference and subsequent species tree analysis. The subfamilies Brachyrhopalinae, Dasypogoninae, Dioctriinae, Stenopogoninae, Tillobromatinae, Trigonomiminae, and Willistonininae were not recovered as monophyletic in either analysis, consistent with a previous molecular study. The inter-subfamily relationships are summarized as follows: Laphriinae and Dioctriinae (in part) are successively sister to the remaining subfamilies, which form two clades; the first consists of a grade of Stenopogoninae (in part), Willistonininae (in part), Bathypogoninae+Phellinae, Stichopogoninae, Leptogastrinae, Ommatiinae, and Asilinae; the second clade consists of a thoroughly paraphyletic assemblage of genera from Dioctriinae (in part), Trigonomiminae, Stenopogoninae (in part), Tillobromatinae, Brachyrhopalinae, and Dasypogoninae. We find that nodal support does not significantly vary with missing data. Furthermore, the bootstrap appears to overestimate nodal support, as has been reported from many recent studies. Gene concordance and site concordance factors seem to perform better, but may actually underestimate support. We instead recommend quartet concordance as a more appropriate estimator of nodal support. Our comprehensive phylogeny demonstrates that the higher classification of Asilidae is far from settled, and it will provide a much-needed foundation for a thorough revision of the subfamily classification.