Mutually exclusive phylogenomic inferences at the root of the angiosperms: Amborella is supported as sister and Observed Variability is biased.

ABSTRACT

Identifying the extant sister group to the remaining angiosperms has been a subject of long debate, for which the primary currently competing hypotheses are that Amborella alone is sister or that the clade (Amborella, Nymphaeales) is sister. Both Xi et al. (Syst. Biol., 2014, 63, 919) and Goremykin et al. (Syst. Biol., 2015, 64, 879) identified Amborella as sister in concatenation-based phylogenetic analyses of their 310 nuclear genes and 78 plastid genes, respectively. But after application of Observed Variability-based character subsampling, both papers reported the clade (Amborella, Nymphaeales) as sister. Hence alternative character-sampling strategies may produce highly supported yet mutually exclusive phylogenetic inferences when applied to nuclear and plastid genomic data sets. Edwards et al. (Mol. Phylogenet. Evol., 2016, 94, 447) defended Observed Variability and the (Amborella, Nymphaeales) hypothesis. In this study I respond to Edwards et al.'s (Mol. Phylogenet. Evol., 2016, 94, 447) criticisms of Simmons and Gatesy (Mol. Phylogenet. Evol., 2015, 91, 98) and use Edwards et al.'s (Mol. Phylogenet. Evol., 2016, 94, 447) and Goremykin et al.'s (Syst. Biol., 2015, 64, 879) own data to demonstrate that the best-supported phylogenetic hypothesis is that Amborella alone is sister and that the competing evidence in favour of the (Amborella, Nymphaeales) hypothesis is caused primarily by methodological artifacts (biased character deletion by Observed Variability, MP-EST and STAR generally not being robust to the highly divergent and mis-rooted gene trees that were used).