Speaker
Description
The molecular clock is a statistical tool that we use to infer evolutionary rates and timescales from molecular sequence data, with the use of calibrations. These calibrations can include sequences sampled at different points in time for many organisms. Without calibrations, evolutionary rates and times are jointly unidentifiable and thus are required. Before inferring rates and times, it is necessary to determine whether there is sufficient information in the data for molecular dating. Typically these include root-to-tip regressions and date randomisation tests. However, to overcome the limitations of such methods, a fully Bayesian approach known as BETS was developed (Bayesian Evaluation of Temporal Signal). It compares the Bayes factors of identically parameterised models with and without including sampling information. We explored tree prior specification and various molecular clock models on temporal signal detection, to test sensitivity and specificity of different tree priors and their hyperparameters. Three varying microbial datasets and simulated molecular data were analysed, and we found that highly informative priors inconsistent with the data can result in the incorrect detection of temporal signal. Consequently, we recommend using prior predictive simulations when choosing priors to ensure parameters of interest fall within reasonable expectations - namely the evolutionary rate and the age of the root node. Further, prior sensitivity analyses should be conducted to assess the robustness of the posterior to the prior. Finally, a molecular clock that reasonably describes the evolutionary process should be selected.
| Expedited Notification | Yes, I want to opt-in for Expedited Notification |
|---|
