Abstract
The RelTime approach estimates timetrees from molecular data when evolutionary rates vary from branch to branch. It has been shown to perform well in analyses of simulated and empirical datasets where evolutionary rates vary extensively. RelTime is computationally efficient and scales well with increasing volumes of data. Consequently, it is being used for estimating divergence time from large datasets. Until now, RelTime has been used without a mathematical foundation. Here, we show that a relative rate framework (RRF) with a principle of minimum rate change is the basis of RelTime. Under RRF, we present analytical solutions for estimating relative rates and divergence times. For both real and simulated datasets, RRF produces estimates similar to those from Bayesian analyses, but RRF provides orders of magnitude increases in computational speed. These gains rise with increasing volumes of data. The mathematical foundation and computational efficiency of RRF makes it suitable for analysis not only of molecular sequence datasets, but also evolutionary trees where the branch lengths reflect the amount of non-molecular (e.g., morphological and traits) evolutionary changes.