Bringing Together the Living and Dead: Integrating Extant and Fossil Biodiversity in Evolutionary Studies
Brochu, Christopher .
Empirical exploration of calibration sensitivity in molecular dating methods - when are multiple-calibration methods most effective?
SEQUENCE-BASED methods for estimating divergence time can be misled by problems with individual fossil calibrations used to estimate rates. Rate variation between clades limits the utility of any single calibration, and individual calibrations are based on first appearance data that are younger (to an unknown degree) from origination times. Multiple-calibration methods have been developed to partially ameliate these problems, but their application to clades with well-sampled fossil records indicates situations in which they might yield spurious results. Quartet-based methods are highly sensitive to calibration age - younger calibrations support young divergences, sometimes with confidence intervals that exclude known fossils older than the divergence estimate. Very young calibrations have to be at least doubled in age to “correct” this problem. Divergence time estimates grow progressively older as calibration point age increases, and the oldest calibrations support divergence estimates much older than the fossil record would predict. These can be as much as 800 percent older than estimates derived with younger calibration points, and some can be unreasonably old - for example, possibly extending the alligatorid-crocodylid split into the Ordovician (first appearances for both are in the Late Cretaceous). Methods less reliant on uniform rates and that allow multiple calibrations are less sensitive to this particular issue, but some of them (including nonparametric rate smoothing and penalized likelihood) are sensitive to the symmetry of calibration distribution on a tree - if calibrations are overwhelmingly on one side of a symmetrical tree, divergence estimates on the uncalibrated side will be substantially older than those on the calibrated side, as if the nodes are being pulled around the root toward the calibrations. These suggest strategies for maximizing the effectiveness of these methods - calibration points should be chosen from different parts of a clade's history and as broadly among ingroup taxa as possible.
Log in to add this item to your schedule
1 - University of Iowa, Department of Geoscience, Iowa City, Iowa, 52242, USA
Presentation Type: Symposium or Colloquium Presentation
Location: 134/Performing Arts Center
Date: Tuesday, August 1st, 2006
Time: 4:15 PM