Overview

PrioriTree is an interactive web application (https://jsigao.shinyapps.io/prioritree/) to help researchers set up BEAST phylodynamic analyses to infer biogeographic history (focussed on prior specification and evaluation). Its development is motivated by properties of the discrete-biogeographic models implemented in BEAST (Lemey et al. 2009; Edwards et al. 2011) and issues with the commonly used default priors in BEAST, identified by Gao et al. (2021) (see the paper for detailed discussions).

These biogeographic models contain many parameters that must be inferred from minimal information (the single geographic area in which each pathogen lineage occurs). Biogeographic inference under this approach is inherently sensitive to the assumed priors. Provided a priori information about the pathogen biology and the dispersal history, PrioriTree allows users to incorporate these biologically informed prior believes into the Bayesian discrete-biogeographic phylodynamic inference. Users can directly express prior assumptions on biologically relevant parameters (e.g., the number of pathogen dispersal events over the entire epidemic history); PrioriTree transforms these assumptions, specifying the resulting priors on the model parameters accordingly.

PrioriTree Main Interface. The interface of the program is partitioned into two main columns: left and right. The left panel takes user input, while the right panel renders the resulting prior distributions, methods description, and the <tt>BEAST</tt> XML script dynamically according to the input.

Figure 0.1: PrioriTree Main Interface. The interface of the program is partitioned into two main columns: left and right. The left panel takes user input, while the right panel renders the resulting prior distributions, methods description, and the BEAST XML script dynamically according to the input.

PrioriTree also allows users to specify these prior assumptions in an interactive manner; it dynamically renders the resulting prior distribution according to user specification in real time. Users can configure other settings (e.g., inferring number of dispersal events between each pair of geographic areas) of the BEAST analysis in PrioriTree as well. Users can view the changes to the BEAST XML script and methods description on the fly according to the input. At the end, PrioriTree generates a readily runnable BEAST XML script (as well as the associated methods template) to perform the analysis that the user conceives. The other main functionality provided by PrioriTree is setting up additional analyses to evaluate the impact of alternative discrete-biogeographic (prior)model specification and visualizing the result. These model-exploration analyses include: posterior-predictive checking, data cloning, and robust Bayesian.

This manual consists of two major sections: a quick-start section that walks through basic functions of PrioriTree and the necessary steps to produce a BEAST XML script for a discrete-biogeographic phylodynamic analysis with PrioriTree, and a thorough-guide section that go into the details about model and prior specification as well as various types of model-exploration analyses that can be set up using PrioriTree. Example input files can be found in this downloadable folder. More example files with larger datasets can be found in this supplementary repository for Gao et al. (2021).

References

Edwards, Ceiridwen J, Marc A Suchard, Philippe Lemey, John J Welch, Ian Barnes, Tara L Fulton, Ross Barnett, et al. 2011. “Ancient Hybridization and an Irish Origin for the Modern Polar Bear Matriline.” Current Biology 21 (15): 1251–8.

Gao, Jiansi, Michael R May, Bruce Rannala, and Brian R Moore. 2021. “The Impact of Prior Misspecification on Bayesian Phylodynamic Inference of Biogeographic History.” bioRxiv.

Lemey, Philippe, Andrew Rambaut, Alexei J Drummond, and Marc A Suchard. 2009. “Bayesian Phylogeography Finds Its Roots.” PLoS Computational Biology 5 (9): e1000520.