Principles, error sources and application suggestions of prevailing molecular dating methods

Abstract

Background & Aims: Molecular dating methods have been applied widely in recent years and provide an indispensable and detailed evolutionary timescale for macroevolutionary researches, particularly for studies on the evolutionary history of biodiversity patterns. Bayesian methods and Markov chain Monte Carlo methods can accommodate multi-dimensional and various type of data and parameter settings, which have helped the node-dating methods implemented in softwares such as BEAST, PAML-MCMCTree to become the most widely used molecular dating methods. One of the advantages of Bayesian frameworks is that they can employ complex models to consider a variety of uncertainty factors to make more accurate estimations of evolutionary divergence times. Progress: We review the principles and main types of Bayesian molecular dating methods and use Bayesian node-dating methods as an example to discuss potential errors in molecular clock models, selection and placement of fossil calibrating points, frequency of sampling, and setting a prior distribution for node calibrations based on fossils. We further describe advantages associated with different Bayesian time tree reconstruction software packages, the discussing principle of node age, and the comparison method of time tree under different models. We also provide suggestions for overcoming the challenges of overestimation and underestimation bias of node ages. Integration of the output of various Bayesian methods and models and selection of the best among them often improve the reliability of molecular dating results. Prospect: Researchers should explicitly assess the relationship between model output of time tree construction and model parameter settings, which increases transparency and provides documentation and reference for future researches. We recommend that future research simultaneously focus on updating fossil records and improving molecular dating methods.