Exploration of data partitioning in an eight-gene data set : Phylogeny of metalmark moths (Lepidoptera, Choreutidae)
Molecular data sets for phylogenetic inference continue to increase in size, especially with respect to the number of genes sampled. As more and more genes are included in analyses, the importance of partitioning the data to avoid problems that can arise from underparameterization becomes more apparent. With an eight-gene data set from 38 metalmark moth species (12 genera represented) and three outgroups, we explored different data partitioning strategies and their influence on convergence and mixing of Markov Chains Monte Carlo in a Bayesian setting. We found that in larger data sets, with an increase in the number of partitions that are made a priori (e.g. by gene and codon position), convergence and mixing become poor. This problem can be overcome by using a recently published algorithm in which homologous sites are grouped into blocks with similar evolutionary rates that can then be modelled as separate data subsets. Using this novel approach to data partitioning, our analyses resolve with strong support relationships among the genera of metalmark moths. Support for the monophyly of the family, the two subfamilies and all genera except Hemerophila is strong. Hemerophila is broken into two separate clades, Hemerophila sensu stricto and another well-supported clade. To render Hemerophila monophyletic, we describe a new genus, Ornarantia Rota, gen. nov., and transfer 18 species from Hemerophila to it. The type species of Ornarantia is Hemerophila laciniosella Busck, 1914.
- ISSN: 0300-3256