Rxivist logo

A Posteriori Evaluation Of Molecular Divergence Dates Using Empirical Estimates Of Time-Heterogeneous Fossilization Rates

By Simon Gunkel, Jes Rust, Torsten Wappler, Christoph Mayer, Oliver Niehuis, Bernhard Misof

Posted 18 Apr 2017
bioRxiv DOI: 10.1101/128314

The application of molecular clock concepts in phylogenetics permits estimating the divergence times of clades with an incomplete fossil record. However, the reliability of this approach is disputed, because the resulting estimates are often inconsistent with different sets of fossils and other parameters (clock models and prior settings) in the analyses. Here, we present the λ statistic, a likelihood approach for a posteriori evaluating the reliability of estimated divergence times. The λ statistic is based on empirically derived fossilization rates and evaluates the fit of estimated divergence times to the fossil record. We tested the performance of this measure with simulated data sets. Furthermore, we applied it to the estimated divergence times of (i) Clavigeritae beetles of the family Staphylinidae and (ii) all extant insect orders. The reanalyzed beetle data supports the originally published results, but shows that several fossil calibrations used do not increase the reliability of the divergence time estimates. Analyses of estimated inter-ordinal insect divergences indicate that uniform priors with soft bounds marginally outperform log-normal priors on node ages. Furthermore, a posteriori evaluation of the original published analysis indicates that several inter-ordinal divergence estimates might be too young. The λ statistic allows the comparative evaluation of any clade divergence estimate derived from different calibration approaches. Consequently, the application of different algorithms, software tools, and calibration schemes can be empirically assessed.

Download data

  • Downloaded 659 times
  • Download rankings, all-time:
    • Site-wide: 51,481
    • In paleontology: 90
  • Year to date:
    • Site-wide: 138,592
  • Since beginning of last month:
    • Site-wide: 120,634

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide