Monday, June 28, 2010

Calibrating Phylogenies from the Fossil Record

Still at "Evolution 2010," we just saw a great talk by Charles Marshall, formerly at Harvard but now at UC Berkeley, about calibrating molecular phylogenies using the fossil records. This is a much more complicated problem that it would seem at first glance because: 1) the maximum likelihood estimate of a node age based on a fossil series from the descendant lineage is biased (towards the present); 2) an assumption that fossilization is temporally random allows for a simple correction to the maximum likelihood estimate - but this assumption is (invariably) violated in empirical data; 3) fossils are not only temporally, but also geographically, non-random; and, finally, 4) the rock record is globally incomplete in some geological eras. Charles's talk included some fantastic graphical simulations of sedimentation and "de-sedimentation" (the removal of previously deposited sediments) as ocean sea levels rise and fall over geological timescales. Evidently, though, in spite of these significant complications, there is still hope that the use of fossil calibrations can improve molecular phylogenetic estimates of species divergence dates. As usual Charles was an intensively engaging speaker and gave a great seminar!


Jim Schulte said...


I think "desedimentation" is better known by it's other name, "erosion".

Very good but kind of sobering talk..


Glor said...

Definitely a sobering talk. The type of information Charles discussed is only available for a very small number of taxa.