Estimating divergence times in large molecular phylogenies

TLDR

Molecular dating adds temporal context to the Tree of Life, yet large, diverse datasets are hampered by prohibitive computation times, complex rate heterogeneity, and scarce reliable clock calibrations. The study introduces a method that estimates relative divergence times for all nodes in very large phylogenetic trees without assuming a specific rate model or clock calibrations. RelTime computes these relative times by applying a model‑free algorithm that bypasses explicit rate variation modeling. RelTime outperformed existing methods on simulated large datasets, achieving up to 1,000‑fold speed gains over the fastest Bayesian approach while maintaining accuracy, thereby enabling efficient dating of very large phylogenies.

Abstract

Molecular dating of species divergences has become an important means to add a temporal dimension to the Tree of Life. Increasingly larger datasets encompassing greater taxonomic diversity are becoming available to generate molecular timetrees by using sophisticated methods that model rate variation among lineages. However, the practical application of these methods is challenging because of the exorbitant calculation times required by current methods for contemporary data sizes, the difficulty in correctly modeling the rate heterogeneity in highly diverse taxonomic groups, and the lack of reliable clock calibrations and their uncertainty distributions for most groups of species. Here, we present a method that estimates relative times of divergences for all branching points (nodes) in very large phylogenetic trees without assuming a specific model for lineage rate variation or specifying any clock calibrations. The method (RelTime) performed better than existing methods when applied to very large computer simulated datasets where evolutionary rates were varied extensively among lineages by following autocorrelated and uncorrelated models. On average, RelTime completed calculations 1,000 times faster than the fastest Bayesian method, with even greater speed difference for larger number of sequences. This speed and accuracy will enable molecular dating analysis of very large datasets. Relative time estimates will be useful for determining the relative ordering and spacing of speciation events, identifying lineages with significantly slower or faster evolutionary rates, diagnosing the effect of selected calibrations on absolute divergence times, and estimating absolute times of divergence when highly reliable calibration points are available.