Bayesian inference of ancestral dates on bacterial phylogenetic trees.
Didelot X, Croucher NJ, Bentley SD, Harris SR, Wilson DJ. · 2018
Read Bayesian inference of ancestral dates on bacterial phylogenetic trees. by Didelot X, Croucher NJ, Bentley SD, Harris SR, Wilson DJ. (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6294524.txt.
Introduction
Didelot and co-authors present a faster Bayesian method for dating bacterial family trees. The method appeared in Nucleic Acids Research. It is packaged as an R tool called BactDating. Dated trees show when shared ancestors of strains likely lived. Such dates can reveal when a key lineage first emerged.
Older tools built the tree and its dates in one slow step. BactDating instead takes a tree built by standard methods. It then infers the ancestor dates on that fixed tree. This design also lets it handle bacterial recombination. The team tested the method on simulated data sets. They also applied it to three real bacterial studies. Results support its speed and accuracy.
Catalog record
| CREATOR | Didelot X, Croucher NJ, Bentley SD, Harris SR, Wilson DJ. |
|---|---|
| YEAR | 2018 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Bayesian inference of ancestral dates on bacterial phylogenetic trees By Xavier Didelot, Nicholas J Croucher, Stephen D Bentley, Simon R Harris, Daniel J Wilson Published in: Nucleic Acids Research DOI: 10.1093/nar/gky783 ABSTRACT Abstract The sequencing and comparative analysis of a collection of bacterial genomes from a single species or lineage of interest can lead to key insights into its evolution, ecology or epidemiology. The tool of choice for such a study is often to build a phylogenetic tree, and more specifically when possible a dated phylogeny, in which the dates of all common ancestors are estimated. Here, we propose a new Bayesian methodology to construct dated phylogenies which is specifically designed for bacterial genomics. Unlike previous Bayesian methods aimed at building dated phylogenies, we consider that the phylogenetic relationships between the genomes have been previously evaluated using a standard phylogenetic method, which makes our methodology much faster and scalable. This two-step approach also allows us to directly exploit existing phylogenetic methods that detect bacterial recombination, and therefore to account for the effect of recombination in the construction of a dated phylogeny. We analysed many simulated datasets in order to benchmark the performance of our approach in a wide range of situations. Furthermore, we present applications to three different real datasets from recent bacterial genomic studies. Our methodology is implemented in a R package called BactDating which is freely available for download at https://github.com/xavierdidelot/BactDating. INTRODUCTION A population evolving sufficiently quickly over a sufficiently long sampling time frame is said to be ‘measurably evolving’, which means that it is possible to estimate the rates over time at which evolution operates and the dates at which ancestors existed (1). This concept has recently become applicable to bacterial species, following the advent of whole-genome sequencing data, in which the relatively low per site evolutionary rates in bacteria are compensated by long genomes, typically comprising millions of sites (2). Consequently, analytical methods that were previously the hallmark of viral genetics are growing in popularity in bacterial genetics, especially the estimation of dated genealogies through the application of the software BEAST (3–6). In a dated phylogeny (also sometimes known as a time-stamped phylogeny or time-calibrated phylogeny), the branch lengths are measured in unit of time (for example days or years), the leaves are shown at known dates of isolation, and the internal nodes are represented at the dates when common ancestors are estimated to have existed. Such estimation of ancestral dates can often provide direct biological insights, for example to date the emergence of an epidemiologically important lineage, but can also be used as a starting point for further analysis, for example to infer past population size dynamics (7), to reconstruct transmission events between hosts (8), to estimate the parameters of an epidemiological model (9), to investigate geographical range expansion (10) or to study ecological adaptation to host species (11). The BEAST framework is popular because it includes many models and extensions, and is based on the Bayesian paradigm which enables a complete quantification of uncertainties in date estimates. However, it is sometimes too slow and computationally demanding to be used, especially when large numbers of sequences are involved. Alternatives based on optimization have therefore started to appear, including LSD (12) which uses least-square optimization methods and TempEst (13) which uses a linear regression to explore the temporal structure of the data. A systematic comparison between LSD, TempEst and BEAST reported that they produced highly congruent estimates of evolutionary rates (14). More recently, three new optimization methods have been released based on maximum likelihood, namely node.dating (15), treedater (16) and TreeTime (17). All these methods are faster than BEAST and able to deal with larger datasets, in great part due to the fact that they assume that phylogenetic relationships have previously been assessed. Their input data therefore consists of the sampling dates plus an unrooted phylogenetic tree which needs to be built in a separate analytical step using a standard phylogenetic software such as RAxML (18), PhyML (19), FastTree (20) or IQ-TREE (21). Here, we present a new methodology called BactDating for analyzing dated genetic data in order to estimate evolutionary rates and dated phylogenies in bacterial populations. We use a Bayesian framework for inference as in BEAST, but consider that phylogenetic relationships have been assessed in a previous step as in the optimization and maximum likelihood methods described above. This way we enjoy the benefits of Bayesian inference in ancestor dating (22), such as assessment of uncertainties and flexibility of model choice and comparison, but with a computational scalability and speed comparable to the optimization methods described above. Furthermore, we explore the specific problems posed by application in bacterial genomics, and in particular the disruptive effect that homologous recombination can have on estimates of the temporal signal (23,24). Recombination is well known for disrupting phylogenetic inference, and especially to affect branch lengths estimates so that trees look star-like with abnormally long terminal branches (23,25,26). To account for this, sites detected as recombinant are sometimes removed prior to running BEAST, but this approach is inefficient and can even exacerbate the problem (23). A more principled method is implemented in the Bacter package (27) which incorporates the ClonalOrigin model of bacterial recombination (28) within BEAST2 (5), but this approach is too computationally intensive to be applicable to large genomic datasets. Instead we show how the effect of recombination can be accounted for in the dating of ancestral nodes, by exploiting a scalable phylogenetic method that accounts for bacterial recombination such as ClonalFrameML (29) or Gubbins (30). We applied BactDating to a large number of datasets simulated under various conditions in order to benchmark its ability to produce correct estimates by comparison with the correct parameter values used during simulation. We also demonstrate the usefulness of BactDating on three case studies based on real datasets from recently published bacterial genomic studies. The first case study used ancient DNA sequencing in order to compare medieval and modern genomes of the leprosy causing pathogen Mycobacterium leprae (31). In the second case study a large number of isolates from clonal lineage of Shigella sonnei from Vietnam were sequenced and compared to study local emergence and dissemination (32). Finally, in the third case study, a worldwide collection of genomes from a highly recombining lineage of Streptococcus pneumoniae were used to investigate its global success and spread (33). MATERIALS AND METHODS Overview of Bayesian inference We consider as input a phylogenetic tree \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\mathcal {P}$\end{document} previously estimated from a set of n bacterial genomes using a standard phylogenetic method. For ease of presentation, we initially make two simplifying assumptions that will be relaxed later. Firstly, we consider that all the isolation dates of the genomes are known. Secondly, we assume that the tree \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\mathcal {P}$\end{document} is already rooted, so that it contains b = 2n − 2 branches. Our aim is to estimate a dated tree \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\mathcal {T}$\end{document}, which in this case means estimating the dates at which each of the n − 1 internal nodes in \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek}