Assessing the relative performance of fast molecular dating methods for phylogenomic data

Abstract

Advances in genome sequencing techniques produced a significant growth of phylogenomic datasets. This massive amount of data represents a computational challenge for molecular dating with Bayesian approaches. Rapid molecular dating methods have been proposed over the last few decades to overcome these issues. However, a comparative evaluation of their relative performance on empirical data sets is lacking. We analyzed 23 empirical phylogenomic datasets to investigate the performance of two commonly employed fast dating methodologies: penalized likelihood (PL), implemented in treePL, and the relative rate framework (RRF), implemented in RelTime. They were compared to Bayesian analyses using the closest possible substitution models and calibration settings. We found that RRF was computationally faster and generally provided node age estimates statistically equivalent to Bayesian divergence times. PL time estimates consistently exhibited low levels of uncertainty. Overall, to approximate Bayesian approaches, RelTime is an efficient method with significantly lower computational demand, being more than 100 times faster than treePL. Thus, to alleviate the computational burden of Bayesian divergence time inference in the era of massive genomic data, molecular dating can be facilitated using the RRF, allowing evolutionary hypotheses to be tested more quickly and efficiently.

Speciation across the Earth driven by global cooling in terrestrial orchids

“…Our tree is largely consilient with other recent estimates ( 19 , 22 ). We used 15 secondary calibrations from previously inferred divergence estimates ( 19 ) to time-calibrate our tree with an uncorrelated relaxed clock in RelTime ( 23 ), which performs well in large empirical datasets ( 24 , 25 ). The 15 secondary calibrations were themselves derived from plastome data and fossil constraints sampled across monocot orders in a Bayesian framework, using an uncorrelated, relaxed, lognormal clock ( 19 ).…”

“…Though the distributions of correlation coefficients were significantly different from zero, both atmospheric CO 2 and sea level were more weakly correlated with speciation than that with temperature (average of DCCA correlation coefficients: CO 2 = 0.29, P < 0.0001; sea level = −0.25, P < 0.0001). To assess potential for method artifacts of time calibration, we infer an alternative framework under penalized likelihood, which is known to perform more poorly and estimate different node ages than RelTime ( 23 – 25 ). Despite different node ages (t 1473 = −61.4, P < 0.0001), we find that the node ages of RelTime and treePL frameworks are highly correlated (r = 0.97, P < 0.0001), and that both frameworks show strong relationships between global cooling and speciation rate ( SI Appendix , Fig.…”

“… 20 ), in order to improve the likelihood calculation, as in other large phylogenies ( 105 ). Because large datasets pose a computational burden for molecular dating methods that rely on Bayesian MCMC sampling, we calibrated the ML phylogeny against geological time using the relaxed-clock ML method RelTime ( 23 – 25 ). Orchids are poorly represented in the fossil record, with only one orchidoid fossil assigned with certainty [to subtribe Goodyerinae ( 106 , 107 )].…”

Phylogenomics reveals the deep ocean as an accelerator for evolutionary diversification in anglerfishes

“…RelTime does not use priors on lineage rates, and instead computes relative time and lineage rates directly from branch lengths in the phylogram (the "relative rate framework") 80,81 . Note that RelTime tends to underestimate divergence times for branches with very few molecular substitutions, unlike methods that include a tree prior 149,150 .…”

Comparative phylogenomic analyses of SNP versus full locus datasets: insights and recommendations for researchers

“…Due to the large number of taxa incorporated into phylogenetic trees in the genomics era, rate smoothing approaches are often used to convert the best tree from a maximum likelihood analysis or maximum clade credibility Bayesian framework to an ultrametric tree (Ho and Duchêne, 2014;Barba-Montoya et al, 2021;Costa et al, 2022). RelTime, which is based on the relative rate framework (RRF) (Tamura et al, 2012(Tamura et al, , 2018 and implemented in MEGA11 (Tamura et al, 2021), was used for divergence time estimation with the sequence alignments and RAxML trees as input.…”