This is an archived static version of the original discussion site.

RelTime rates collapses to a strict clock


Phylobabble readers, especially those considering using the dating program RelTime, might be interested in this paper in press at GBE:

RelTime rates collapses to a strict clock when estimating the timeline of animal diversification

Establishing an accurate timescale for the history of life is crucial to understand evolutionary processes. To this scope, relaxed molecular clock models implemented in a Bayesian MCMC framework are generally used. However, these methods are time consuming. RelTime, a non-Bayesian method implementing a fast, ad hoc, algorithm for relative dating, was developed to overcome the computational inefficiencies of Bayesian software. RelTime was recently used to investigate the origin of animals, and found results consistent with outdated studies from the 1980es and 1990es that recovered metazoans to have a Mesoproterozoic origin (originating approximately 1.5 Billion years ago) based on a strict clock model. RelTime results are unexpected and disagree with the largest majority of modern, relaxed, Bayesian molecular clock analyses, that suggest animals originated in the Tonian-Cryogenian (less than 850 millions of years ago) instead. Here, we demonstrate that RelTime-inferred divergence times for the origin of animals are erroneous, a consequence of the inability of RelTime to relax the clock along the internal branches of the animal phylogeny. RelTime-inferred divergence times are comparable to strict-clock estimates because they are essentially inferred under a strict clock. Our results warn us of the danger of using ad hoc algorithms making implicit, unclear, assumptions about rate changes along a tree. Our study reject the existence of Mesoproterozioc animals. Metazoans emerged in the Tonian-Cryogenian, and diversified in the Ediacaran, in the immediate prelude to the routine fossilization of animals in the Cambrian associated with the emergence of readily preserved skeletons.


Very interested to know if others here have used / compared RelTime on empirical or simulated data for comparing relative rates. The simulations in the original ms are (of course!!!) encouraging. But the ad-hoc nature of the method is a worry, and this surely can’t be the first time that someone’s come across interesting issues.