TWO NEW APPROACHES TO CALIBRATING MOLECULAR CLOCKS AND EVALUATION OF BIAS IN EXISTING METHODS

 

Ryan W. Norris, Cory L. Strope, Arlin Stoltzfus

 

Modern molecular clock computer programs provide many options for the shape of the prior distribution at a node that is calibrated based on fossil data, but very little work has been done to determine which of these parameters are appropriate given a particular set of fossil data.  We introduce two new methods for generating priors at calibrating nodes based on the date of the oldest fossil added to either the difference between the two oldest fossils (the penultimate gap) or the difference between sister lineages (the ghost lineage length) divided by two. Using a set of trees with simulated branch lengths and with nucleotides and fossil dates generated using a modified version of indel-Seq-Gen, we test the penultimate gap (PenG) and ghost lineage length (GLin) methods against existing methods.  The results of our simulations indicate that PenG and GLin are relatively unbiased estimators that produce results that are both accurate and precise compared to existing methods.