Effects of taxon sampling on molecular dating for within-genus divergence events, when deep fossils are used for calibration

摘要

A universal method of molecular dating that can be applied to all families and genera regardless of their fossil records, or lack thereof, is highly desirable. A possible method for eudicots is to use a large phylogeny calibrated using deep fossils including tricolpate pollen as a fixed (124 mya) calibration point. This method was used to calculate node ages within three species-poor disjunct basal eudicot genera, Caulophyllum, Podophyllum and Pachysandra, and sensitivity of these ages to effects such as taxon sampling were then quantified. By deleting from one to three accessions related to each genus in 112 different combinations, a confidence range describing variation due only to taxon sampling was generated. Ranges for Caulophyllum, Podophyllum and Pachysandra were 8.4–10.6, 7.6–20.0, and 17.6–25.0 mya, respectively. However, the confidence ranges calculated using bootstrapping were much wider, at 3–19, 0–32 and 11–32 mya, respectively. Furthermore, deleting 10 adjacent taxa had a large effect in Pachysandra only, indicating that undersampling effects are significant among Buxales. Changes to sampling density in neighboring clades, or to the position of the fixed fossil calibration point had small to negligible effects. Non-parametric rate smoothing was more sensitive to taxon sampling effects than was penalized likelihood. The wide range for Podophyllum, compared to the other two genera, was probably due to a high degree of rate heterogeneity within this genus. Confidence ranges calculated by this method could be narrowed by sampling more individuals within the genus of interest, and by sequencing multiple DNA regions from all species in the phylogeny.