Written and maintained by CASRAI Editorial Board
Last updated
BEAST X is open-source software for Bayesian phylogenetic, phylogeographic and phylodynamic inference: it estimates evolutionary trees, and the parameters of models fitted to them, from molecular sequence data using Markov chain Monte Carlo (MCMC). This page is about the software, not the car, bat or comic character that share the name.
Short answer
BEAST X (version 10.5.0 at the time of writing) is a cross-platform program for Bayesian analysis of molecular sequences, centred on rooted, time-measured phylogenies under strict or relaxed molecular clock models. Rather than reporting one best tree, it uses MCMC to average over tree space, weighting trees by posterior probability. It is described in Baele et al. (2025), Nature Methods 22(8):1653-1656.
Definition and scope
The BEAST project site describes the program as a cross-platform tool for Bayesian analysis of molecular sequences with MCMC. The 2025 paper presents BEAST X as an open-source, cross-platform release that adds new model families and faster inference. Its source code is on GitHub under the LGPL licence, and it needs Java 1.8 or later. The authors present it as related to, but independent of, BEAST 2, which the project site describes as an independent project led by the University of Auckland; the two should be treated as separate programs with their own documentation.
The software is more than one executable. BEAUti is the interface used to set up an analysis, the BEAST engine runs the MCMC chain, and companion programs such as Tracer and FigTree, listed on the project site, are used afterwards to inspect and view the output.
How it is used in practice
The project site provides installation guides for Mac, Windows and UNIX/Linux, and tutorials with example files. Many of the new models in BEAST X use Hamiltonian Monte Carlo, a sampler that needs gradients of the likelihood. The paper reports fast approximate likelihood gradients for all unknown substitution-model parameters and preorder tree traversals that make gradient computation linear in the number of taxa. It reports gains in effective sample size (ESS) per unit time over univariate samplers, with the size of the speedup depending on dataset size and tuning.
The program works with the BEAGLE library (version 4.0 or later) to run likelihood calculations on multicore CPUs and GPUs. As a worked example, the paper analyses the invasion of Omicron BA.1 in England using about 11,351 genomes, 256 lower-tier local authorities and roughly 22,700 branch-specific rate parameters; the XML inputs are on GitHub.
What BEAST X adds, by model family
| Model family | Additions described in Baele et al. (2025) |
|---|---|
| Substitution models | Markov-modulated models and random-effects models that relax reversibility |
| Molecular clocks | Time-dependent rates, continuous and mixed-effects relaxed clocks, shrinkage-based random local clock |
| Phylogeography | Generalised linear model extensions that handle missing predictors, and relaxed random walks |
| Trait evolution | Gaussian models including Ornstein-Uhlenbeck, missing-trait models, phylogenetic factor analysis, multivariate probit |
| Demography | Nonparametric skygrid, preferential-sampling correction, episodic birth-death sampling |
Where it matters for labs and computing support
For a lab or research-computing team, the sourced requirements form a short pre-run checklist.
- Java 1.8 or later is installed (paper).
- BEAGLE 4.0 or later is available if the analysis is to use multicore CPU or GPU offload (paper).
- The analysis is configured in BEAUti, the set-up interface (project site).
- Output is inspected in Tracer before results are interpreted (the project site lists Tracer as an analysis program); see the guide to Markov chain Monte Carlo for what convergence diagnostics such as ESS mean.
Once a tree is estimated, how to read a phylogenetic tree explains the output, and making a phylogenetic tree places Bayesian methods alongside other approaches. For a maximum-likelihood alternative see IQ-TREE, and for plotting trees in R see ggtree. The wider field is introduced in what is evolutionary biology.
Common misunderstandings
Is BEAST X the same as BEAST 2?
No. The project site describes BEAST 2 as an independent project led by the University of Auckland, and the BEAST X authors present their software as related but independent. Check each program’s own documentation before assuming that settings or files carry across.
Does it return a single best tree?
No. It uses MCMC to average over tree space, weighting trees by posterior probability, so results are distributions over trees and parameters rather than one point estimate.
Is a GPU required?
The paper describes BEAGLE as supporting multicore CPU and GPU computation; it does not describe a GPU as mandatory.
FAQ
What is BEAUti?
BEAUti is the user-interface program the project site describes as simple to use for setting up BEAST analyses.
Is BEAST X free?
The paper describes it as open source under the LGPL licence, with source code on GitHub.
Which operating systems are supported?
The project site provides install guides for Mac, Windows and UNIX/Linux, and the paper describes the software as cross-platform.
Sources
- BEAST project site, beast.community, read 2026-11-08.
- Baele G, Ji X, Hassler GW, McCrone JT, Shao Y, Zhang Z, Holbrook AJ, Lemey P, Drummond AJ, Rambaut A, Suchard MA. BEAST X for Bayesian phylogenetic, phylogeographic and phylodynamic inference. Nature Methods 22(8):1653-1656 (2025). doi:10.1038/s41592-025-02751-x; open-access text at PubMed Central, read 2026-11-08.
