Introduction to Bayesian phylogenetics and BEAST
Download as PPTX, PDF3 likes2,642 views
This document provides an overview of a course on Bayesian phylogenetics and the BEAST software package. The course covers introductory topics on Bayesian analysis and BEAST, as well as more advanced analyses including incorporating temporal and trait data. The document outlines the organization and topics to be covered in lectures, including why Bayesian methods are well-suited for pathogen evolution analysis and an introduction to Markov chain Monte Carlo sampling. It also provides information on setting up BEAST analyses using BEAUti, evaluating runs in Tracer, and summarizing runs using LogCombiner and TreeAnnotator.
![Setting up the analysis: BEAUTi
Import data: Nexus format
#NEXUS
[These are comments.
They are ignored by the program.]
Begin data;
dimensions ntax=5 nchar=15;
format datatype=DNA gap=- missing=?;
matrix
Bug1 ACCTGATTACGGGCA
Bug2 ACCCGAATACGGACA
Bug3 ACCTATTTACGCCCA
BugF ACTATATTACCGGCA
BugBX4W ACCAAA---CGGGCA
;
End;](https://image.slidesharecdn.com/lecture1feb2016-170302023639/85/Introduction-to-Bayesian-phylogenetics-and-BEAST-16-320.jpg)
![Estimating best-fit models and initial parameters:
jModelTest
Model selected: TVM+I+G
-lnL = 1676.8109
K = 9
AIC = 3371.6218
Base frequencies:
freqA = 0.2259
freqC = 0.3199
freqG = 0.2405
freqT = 0.2137
Substitution model:
Rate matrix
R(a) [A-C] = 0.2494
R(b) [A-G] = 4.8655
R(c) [A-T] = 0.7435
R(d) [C-G] = 0.3907
R(e) [C-T] = 4.8655
R(f) [G-T] = 1.0000
Among-site rate variation
Proportion of invariable sites (I) = 0.6508
Variable sites (G)
Gamma distribution shape parameter = 0.5913
Setting up the analysis: Models](https://image.slidesharecdn.com/lecture1feb2016-170302023639/85/Introduction-to-Bayesian-phylogenetics-and-BEAST-20-320.jpg)
![Running BEAST: swarm on Biowulf
• Requires a .swarm file
• A text file containing
• Run in command line
beast beastJob_1.xml > beastJob_1out.txt
sleep 2; beast beastJob_2.xml > beastJob_2out.txt
sleep 4; beast beastJob_3.xml > beastJob_3out.txt
sleep 6; beast beastJob_4.xml > beastJob_4out.txt
sleep 8; beast beastJob_5.xml > beastJob_5out.txt
[username]$ swarm -f beastInput.swarm --module BEAST](https://image.slidesharecdn.com/lecture1feb2016-170302023639/85/Introduction-to-Bayesian-phylogenetics-and-BEAST-38-320.jpg)
Introduction to Bayesian phylogenetics and BEAST
- 1. Spring 2016 Kurt Wollenberg, PhD Phylogenetics Specialist Bioinformatics and Computational Biosciences Branch Office of Cyber Infrastructure and Computational Biology Molecular Evolutionary Analysis Using BEAST Part 1: Introduction to Bayesian phylogenetics and BEAST
- 2. Course Organization • Introduction to Bayesian phylogenetics • Introduction to BEAST • Building a Bayesian phylogeny • Incorporating sample time in the phylogeny • Estimating demographic parameters • Estimating species trees from gene trees • Estimating ancestral trait states (esp. geography)
- 3. Lecture Organization • Why Bayesian phylogenetics is well-suited to the analysis of pathogen molecular evolution • A short tour of Bayesian MCMC analysis • What is BEAST? An overview of the BEAST package • BEAST Analysis Demo
- 4. What’s so special about pathogens? • Short generation time • Rapid evolution • Genotypes - easy, phenotypes - hard • Large populations • Structured populations • Rigorous temporal sampling of genotypes
- 5. Why use Bayesian methods on pathogens? • Coalescent approach is more appropriate • Can incorporate temporal data • Can incorporate geographical data • Can incorporate host data
- 6. What is Bayesian analysis? • Calculation of the probability of parameters (tree, substitution model) given the data (sequence alignment) • p(θ|D) = (Likelihood x prior)/probability of the data • p(θ|D) = p(D|θ)p(θ)/p(D)
- 7. Exploring the posterior probability distribution Posterior probabilities of trees and parameters are approximated using Markov Chain Monte Carlo (MCMC) sampling Markov Chain: A statement of the probability of moving from one state to another Bayesian Analysis
- 8. What is MCMC? Markov Chain Monte Carlo Markov chain Monte Carlo One link in the chain Choosing a link
- 9. What is MCMC? Markov Chain Monte Carlo: accept or reject? Metropolis-Hastings algorithm PosteriorProbability Topology A Topology B Topology C 20% 48% 32% Accept! Maybe
- 10. What is BEAST? • Bayesian Evolutionary Analysis Sampling Trees • A collection of programs for performing Bayesian MCMC analysis of molecular sequences • Can incorporate sample time information • Can perform a broad range of other evolutionary analyses using sequence data.
- 11. What is BEAST? The Programs: • BEAUti - Creating XML input files • BEAST - MCMC analysis of molecular sequences • Tracer - Viewing MCMC output • LogCombiner - Combining output files • TreeAnnotator - Generate the consensus tree • FigTree - Drawing a tree
- 12. Different types of BEAST analyses • Calculating a Bayesian coalescent phylogeny • Calculating a Time-Stamped Bayesian coalescent • Estimated population dynamics (Bayesian skyline/skyride/skygrid) • Combined gene and species phylogeny estimate (*BEAST) • Phylogeographic analysis (time and location data)
- 13. Defining your analysis • Prior knowledge of tree? • Calibrating nodes? • Substitution model? • Effective population sizes? • What priors to use?
- 14. Setting up the analysis: BEAUTi
- 15. Setting up the analysis: BEAUTi • Import data – Nexus or fasta format • Incorporate known structure - taxa • Substitution model parameters • Strict or relaxed clock? • Tree prior • Substitution model priors • Adjustments from previous runs (operators) • Setting the chain
- 16. Setting up the analysis: BEAUTi Import data: Nexus format #NEXUS [These are comments. They are ignored by the program.] Begin data; dimensions ntax=5 nchar=15; format datatype=DNA gap=- missing=?; matrix Bug1 ACCTGATTACGGGCA Bug2 ACCCGAATACGGACA Bug3 ACCTATTTACGCCCA BugF ACTATATTACCGGCA BugBX4W ACCAAA---CGGGCA ; End;
- 17. Setting up the analysis: BEAUTi Import data: Fasta format >Bug1 ACCTGATTACGGGCA >Bug2 ACCCGAATACGGACA >Bug3 ACCTATTTACGCCCA >BugF ACTATATTACCGGCA >BugBX4W ACCAAA---CGGGCA
- 18. Setting up the analysis: Models Substitution Models • HKY - Unequal base frequencies and transition/transversion rate ratio • Must specify prior and initial estimates for transition/transversion rate ratio • GTR - Unequal base frequencies and each substitution has its own rate parameter • Must specify prior and initial estimates for each substitution rate (relative to C-T rate)
- 19. Site Models • Site heterogeneity models • Gamma • Modeling rate of change using a discrete gamma distribution • Invariant • Percent of non-variable sites in the data Setting up the analysis: Models
- 20. Estimating best-fit models and initial parameters: jModelTest Model selected: TVM+I+G -lnL = 1676.8109 K = 9 AIC = 3371.6218 Base frequencies: freqA = 0.2259 freqC = 0.3199 freqG = 0.2405 freqT = 0.2137 Substitution model: Rate matrix R(a) [A-C] = 0.2494 R(b) [A-G] = 4.8655 R(c) [A-T] = 0.7435 R(d) [C-G] = 0.3907 R(e) [C-T] = 4.8655 R(f) [G-T] = 1.0000 Among-site rate variation Proportion of invariable sites (I) = 0.6508 Variable sites (G) Gamma distribution shape parameter = 0.5913 Setting up the analysis: Models
- 21. Site heterogeneity models: The Gamma Distribution Mean = kθ Shape parameter = θ Coefficient of Variation = 1/√θ Setting up the analysis: Models
- 22. Setting up the analysis: Models Clock Models • Strict clock – same rate for all branches • Relaxed clock – independent rate among branches • Exponential or Lognormal distribution of rates • For contemporaneous data setting a fixed mean substitution rate of 1.0 (uncheck “Estimate”) results in node ages as substitutions per site (MrBayes branch lengths)
- 23. Setting up the analysis: Models Tree Prior • Coalescent • constant size • exponential growth • GMRF Bayesian Skygrid • Speciation • Yule process • Birth-Death • Epidemiology
- 24. Setting up the analysis: Models Testing Models and Priors Path Sampling/Stepping Stone analysis • Estimation of marginal likelihoods under different analysis parameters. • Invoke on MCMC tab in BEAUti. • Separate runs necessary for each changed parameter. • Runs a complete MCMC analysis, then the X PS/SS iterations.
- 25. Setting up the analysis: Models Testing Models and Priors Path Sampling/Stepping Stone analysis
- 26. Setting up the analysis: Models Testing Models and Priors Path Sampling/Stepping Stone analysis log marginal likelihoods Path Sampling Stepping Stone HKY/strict clock -4725.85 -4728.68 HKY+gi/strict -4515.99 -4518.05 HKY+gi/LN relaxed -4436.10 -4438.75 GTR/strict clock -4746.62 -4749.14 GTR+gi/strict -4526.87 -4529.05 GTR+gi/LN relaxed -4548.39 -4551.22
- 27. Setting up the analysis: Models Testing Models and Priors Does the relaxed clock fit the data? ucld.stdev Frequency 0.5 1 1.5 2 2.5 0 50 100 150 200 250 300 350 400
- 28. Setting up the analysis: Models Testing Models and Priors Does the relaxed clock fit the data? 2014_GN.SL_SRDpart1.ucld.stdev Frequency 0 0.5 1 1.5 2 2.5 3 0 50 100 150 200 250 300 350
- 29. Setting up the analysis: Models Testing Models and Priors Does the relaxed clock fit the data? ucld.stdev Frequency 0 0.5 1 1.5 2 2.5 3 0 500 1000 1500 2000 2500
- 30. Running the analysis: BEAST • Load your input file • That’s it
- 31. Evaluating the analysis: Tracer
- 32. Evaluating the analysis: Tracer • Check for convergence • Evaluating ESS values • Viewing behavior of parameter estimates • Examining traces • Extracting parameter estimates and statistics
- 33. Evaluating the analysis: Tracer • What if my analysis didn’t converge? • Can I make multiple simultaneous runs? • Swarm on Biowulf
- 34. Evaluating the analysis: Tracer What if my analysis didn’t converge?
- 35. Evaluating the analysis: Tracer What if my analysis didn’t converge?
- 36. What if my analysis didn’t converge?
- 37. What if my analysis didn’t converge?
- 38. Running BEAST: swarm on Biowulf • Requires a .swarm file • A text file containing • Run in command line beast beastJob_1.xml > beastJob_1out.txt sleep 2; beast beastJob_2.xml > beastJob_2out.txt sleep 4; beast beastJob_3.xml > beastJob_3out.txt sleep 6; beast beastJob_4.xml > beastJob_4out.txt sleep 8; beast beastJob_5.xml > beastJob_5out.txt [username]$ swarm -f beastInput.swarm --module BEAST
- 39. Merging output files: LogCombiner
- 40. Merging output files: LogCombiner • Log files vs Tree files • Selecting files • Specifying burn-in (number of steps or trees) • Specifying subsampling • Specifying output file
- 41. Merging output files: LogCombiner • Burn in?
- 42. Calculating the tree: TreeAnnotator
- 43. Calculating the tree: TreeAnnotator • Burn in? Number of trees or the number of steps. • Tree Type: MCC, Max sum of CC, or target • Node heights: target, mean, or median • Specify input and output files
- 45. Drawing trees: FigTree • Specifying additional values (esp. posterior probabilities) • Tree appearance • Ordering branches • Re-rooting • Exporting graphics
- 47. Setting up the analysis: BEAUTi
- 48. Running the analysis: BEAST
- 49. Evaluating the analysis: Tracer
- 50. Merging output files: LogCombiner
- 51. Calculating the tree: TreeAnnotator
- 53. BEAST2 • Still … Bayesian Evolutionary Analysis Sampling Trees • Modular rewrite of the BEAST software • Various evolutionary analyses performed through a system of independent software packages. • Access software, documentation, etc., through the website beast2.org • Still a few bugs in the system…
- 54. Seminar Follow-Up Site For access to past recordings, handouts, slides visit this site from the NIH network: http://collab.niaid.nih.gov/sites/research/SIG/Bioinformatics/ 54 1. Select a Subject Matter View: • Seminar Details • Handout and Reference Docs • Relevant Links • Seminar Recording Links 2. Select a Topic Recommended Browsers: • IE for Windows, • Safari for Mac (Firefox on a Mac is incompatible with NIH Authentication technology) Login • If prompted to log in use “NIH” in front of your username
- 57. 57 Questions?
- 58. 58 Next? • Time-structured phylogenies • Estimating demographic parameters • GMRF skyride analysis
- 59. 59