Impact of Partition Models on Phylogenetic Inference and Divergence Times of Lampyridae from Mitochondrial Genomes

Mitochondrial genomes are frequently used for phylogenetic inference due to their availability and cost-efficient sequencing. In most mitogenomic phylogenetic analyses, only the two ribosomal RNA and 13 protein coding genes are used. For such multi-locus datasets it has long been established that appropriate data partition models, e.g., partitioning by gene type and/or codon position, are necessary for robust phylogenetic inference. While most studies focused on the impact of partition models on the tree topology, little is known about the impact on divergence time estimation. Furthermore, although modeling among site rate variation within a partition is common practice, the extent of substitution rate variation among partitions is less explored. Here we study the impact of four partition model strategies: (i) no subdivision of the data (uniform), (ii) partitioning by gene, (iii) partitioning by codon position, and (iv) partitioning by both gene and codon position (combined). We explore the impact of partition models on divergence time estimation in fireflies (Coleoptera: Lampyridae). To this end, we sequenced mitochondrial genomes from 22 firefly species from Europe and Central America and combined these with 82 published mitochondrial genomes. Our results represent the most comprehensive time-calibrated phylogeny of fireflies to day. While the partition models had an impact on the inferred tree topology, the divergence times were almost not affected. Furthermore, we observed up to 3-fold substitution rate variation across genes and additionally more than to 10-fold substitution rate variation across codon positions. Our study gives insights into best practices of performing partitioned-data time-calibrated phylogenetic inference from mitochondrial genomes and multi-locus datasets in general.