Bayesian phylogenetic analysis with MCMC algorithms generates an estimate of the posterior distribution of phylogenetic trees in the form of a sample of phylogenetic trees and related parameters. The high dimensionality and non-Euclidean nature of tree space complicates summarizing the central tendency and variance of the posterior distribution in tree space. Here we introduce a new tractable tree distribution and associated point estimator that can be constructed from a posterior sample of trees. Through simulation studies we show that this point estimator performs at least as well and often better than standard methods of producing Bayesian posterior summary trees. We also show that the method of summary that performs best depends on the sample size and dimensionality of the problem in non-trivial ways.

The spread of infectious diseases is shaped by spatial and temporal aspects, such as host population structure or changes in the transmission rate or number of infected individuals over time. These spatiotemporal dynamics are imprinted in the genomes of pathogens and can be recovered from those genomes using phylodynamics methods. However, phylodynamic methods typically quantify either the temporal or spatial transmission dynamics, which leads to unclear biases, as one can potentially not be inferred without the other. Here, we address this challenge by introducing a structured coalescent skyline approach, MASCOT-Skyline, that allows us to jointly infer spatial and temporal transmission dynamics of infectious diseases using Markov chain Monte Carlo inference. To do so, we model the effective population size dynamics in different locations using a non-parametric function, allowing us to approximate a range of population size dynamics. We show, using a range of different viral outbreak datasets, potential issues with phylogeographic methods. We then use these viral datasets to motivate simulations of outbreaks that illuminate the nature of biases present in the different phylogeographic methods. We show that spatial and temporal dynamics should be modeled jointly, even if one seeks to recover just one of the two. Further, we showcase conditions under which we can expect phylogeographic analyses to be biased, particularly different subsampling approaches, as well as provide recommendations on when we can expect them to perform well. We implemented MASCOT-Skyline as part of the open-source software package MASCOT for the Bayesian phylodynamics platform BEAST2.

Assessing convergence of Markov chain Monte Carlo (MCMC) based analyses is crucial but challenging, especially so in high dimensional and complex spaces such as the space of phylogenetic trees (treespace). In practice, it is assumed that the target distribution is the unique stationary distribution of the MCMC and convergence is achieved when samples appear to be stationary. Here we leverage recent advances in computational geometry of the treespace and introduce a method that combines classical statistical techniques and algorithms with geometric properties of the treespace to automatically evaluate and assess practical convergence of phylogenetic MCMC analyses. Our method monitors convergence across multiple MCMC chains and achieves high accuracy in detecting both practical convergence and convergence issues within treespace. Furthermore, our approach is developed to allow for real-time evaluation during the MCMC algorithm run, eliminating any of the chain post-processing steps that are currently required. Our tool therefore improves reliability and efficiency of MCMC based phylogenetic inference methods and makes analyses easier to reproduce and compare. We demonstrate the efficacy of our diagnostic via a well-calibrated simulation study and provide examples of its performance on real data sets. Although our method performs well in practice, a significant part of the underlying treespace probability theory is still missing, which creates an excellent opportunity for future mathematical research in this area.

Relaxed clock models enable estimation of molecular substitution rates across lineages and are widely used in phylogenetics for dating evolutionary divergence times. Under the (uncorrelated) relaxed clock model, tree branches are associated with molecular substitution rates which are independently and identically distributed. In this article we delved into the internal complexities of the relaxed clock model in order to develop efficient MCMC operators for Bayesian phylogenetic inference. We compared three substitution rate parameterisations, introduced an adaptive operator which learns the weights of other operators during MCMC, and we explored how relaxed clock model estimation can benefit from two cutting-edge proposal kernels: the AVMVN and Bactrian kernels. This work has produced an operator scheme that is up to 65 times more efficient at exploring continuous relaxed clock parameters compared with previous setups, depending on the dataset. Finally, we explored variants of the standard narrow exchange operator which are specifically designed for the relaxed clock model. In the most extreme case, this new operator traversed tree space 40% more efficiently than narrow exchange. The methodologies introduced are adaptive and highly effective on short as well as long alignments. The results are available via the open source optimised relaxed clock (ORC) package for BEAST 2 under a GNU licence (https://github.com/jordandouglas/ORC).

Elaboration of Bayesian phylogenetic inference methods has continued at pace in recent years with major new advances in nearly all aspects of the joint modelling of evolutionary data. It is increasingly appreciated that some evolutionary questions can only be adequately answered by combining evidence from multiple independent sources of data, including genome sequences, sampling dates, phenotypic data, radiocarbon dates, fossil occurrences, and biogeographic range information among others. Including all relevant data into a single joint model is very challenging both conceptually and computationally. Advanced computational software packages that allow robust development of compatible (sub-)models which can be composed into a full model hierarchy have played a key role in these developments. Developing such software frameworks is increasingly a major scientific activity in its own right, and comes with specific challenges, from practical software design, development and engineering challenges to statistical and conceptual modelling challenges. BEAST 2 is one such computational software platform, and was first announced over 4 years ago. Here we describe a series of major new developments in the BEAST 2 core platform and model hierarchy that have occurred since the first release of the software, culminating in the recent 2.5 release.

Model-based phylodynamic approaches recently employed generalized linear models (GLMs) to uncover potential predictors of viral spread. Very recently some of these models have allowed both the predictors and their coefficients to be time-dependent. However, these studies mainly focused on predictors that are assumed to be constant through time. Here we inferred the phylodynamics of avian influenza A virus H9N2 isolated in 12 Asian countries and regions under both discrete trait analysis (DTA) and structured coalescent (MASCOT) approaches. Using MASCOT we applied a new time-dependent GLM to uncover the underlying factors behind H9N2 spread. We curated a rich set of time-series predictors including annual international live poultry trade and national poultry production figures. This time-dependent phylodynamic prediction model was compared to commonly employed time-independent alternatives. Additionally the time-dependent MASCOT model allowed for the estimation of viral effective sub-population sizes and their changes through time, and these effective population dynamics within each country were predicted by a GLM. International annual poultry trade is a strongly supported predictor of virus migration rates. There was also strong support for geographic proximity as a predictor of migration rate in all GLMs investigated. In time-dependent MASCOT models, national poultry production was also identified as a predictor of virus genetic diversity through time and this signal was obvious in mainland China. Our application of a recently introduced time-dependent GLM predictors integrated rich time-series data in Bayesian phylodynamic prediction. We demonstrated the contribution of poultry trade and geographic proximity (potentially unheralded wild bird movements) to avian influenza spread in Asia. To gain a better understanding of the drivers of H9N2 spread, we suggest increased surveillance of the H9N2 virus in countries that are currently under-sampled as well as in wild bird populations in the most affected countries.

We present a new open source, extensible and flexible software platform for Bayesian evolutionary analysis called BEAST 2. This software platform is a re-design of the popular BEAST 1 platform to correct structural deficiencies that became evident as the BEAST 1 software evolved. Key among those deficiencies was the lack of post-deployment extensibility. BEAST 2 now has a fully developed package management system that allows third party developers to write additional functionality that can be directly installed to the BEAST 2 analysis platform via a package manager without requiring a new software release of the platform. This package architecture is showcased with a number of recently published new models encompassing birth-death-sampling tree priors, phylodynamics and model averaging for substitution models and site partitioning. A second major improvement is the ability to read/write the entire state of the MCMC chain to/from disk allowing it to be easily shared between multiple instances of the BEAST software. This facilitates checkpointing and better support for multi-processor and high-end computing extensions. Finally, the functionality in new packages can be easily added to the user interface (BEAUti 2) by a simple XML template-based mechanism because BEAST 2 has been re-designed to provide greater integration between the analysis engine and the user interface so that, for example BEAST and BEAUti use exactly the same XML file format.

MOTIVATION: Bayesian analysis through programs like BEAST (Drummond and Rumbaut, 2007) and MrBayes (Huelsenbeck et al., 2001) provides a powerful method for reconstruction of evolutionary relationships. One of the benefits of Bayesian methods is that well-founded estimates of uncertainty in models can be made available. So, for example, not only the mean time of a most recent common ancestor (tMRCA) is estimated, but also the spread. This distribution over model space is represented by a set of trees, which can be rather large and difficult to interpret. DensiTree is a tool that helps navigating these sets of trees. RESULTS: The main idea behind DensiTree is to draw all trees in the set transparently. As a result, areas where a lot of the trees agree in topology and branch lengths show up as highly colored areas, while areas with little agreement show up as webs. This makes it possible to quickly get an impression of properties of the tree set such as well-supported clades, distribution of tMRCA and areas of topological uncertainty. Thus, DensiTree provides a quick method for qualitative analysis of tree sets. AVAILABILITY: DensiTree is freely available from http://compevol.auckland.ac.nz/software/DensiTree/. The program is licensed under GPL and source code is available. CONTACT: [email protected]

Replicability of machine learning experiments measures how likely it is that the outcome of one experiment is repeated when performed with a different randomization of the data. In this paper, we present an estimator of replicability of an experiment that is efficient. More precisely, the estimator is unbiased and has lowest variance in the class of estimators formed by a linear combination of outcomes of experiments on a given data set.We gathered empirical data for comparing experiments consisting of different sampling schemes and hypothesis tests. Both factors are shown to have an impact on replicability of experiments. The data suggests that sign tests should not be used due to low replicability. Ranked sum tests show better performance, but the combination of a sorted runs sampling scheme with a t-test gives the most desirable performance judged on Type I and II error and replicability.