Rui Borges

Tutorial 2

Bayesian phylogenetic inference

In this tutorial, we will perform phylogenetic analysis with the Metzker et al. (2002) dataset, which includes HIV env/gp120 sequences. This data was used as evidence in a Louisiana gastroenterologist’s trial accused of deliberately infecting someone (the victim) with HIV-infected blood from one of the gastroenterologist’s patients. The jury was charged with addressing whether it was beyond a reasonable doubt that the victim was infected as a result of the gastroenterologist’s actions.

In this exercise, we want to determine if there is evidence that the victim (whose sequences all begin with a V) was directly infected by blood taken from the patient (whose sequences all begin with a P). Therefore, the phylogenetic problem is to establish whether the virus samples in the victim and the patient were closely related. HIV-infected individuals from the local metropolitan area (tagged as LA) were also included as controls. Two more divergent reference sequences were also added: M62320_HIVU455 and K03454_HIVELI. Notice that the sequences used in this tutorial are only a small subset (30 sequences) of the evidence they considered when trying to answer this question.

Sequence alignment: ↘HIV_env_gp120.nxs

We will be performing phylogenetic inference using the MrBayes, which is a software dedicated to reconstructing phylogenetic relationships (among other analyses) between molecular sequences using Bayesian inference. Reference to this method: Huelsenbeck and Ronquist MRBAYES: Bayesian inference of phylogenetic trees Bioinformatics, 17:754-755.

MrBayes can be executed by simply typing mb in the command line. The first thing to do is upload the alignment file and set the outgroup taxa. You can set the outgroup by either writing the sequence name (either M62320_HIVU455 or K03454_HIVELI) or its position on the multiple sequence alignment.

execute HIV_env_g120.nxs
outgroup 30

Next, we define the substitution model. We will use the GTR substitution model with site heterogeneity (following the results in the paper). The models of evolution available in MrBayes are defined under the option nst while the gamma rates can be set in the rates option.

lset nst=6 rates=gamma

As we perform Bayesian inference, we will be running the MCMC algorithm. Set the MCMC to run for 100 000 generations using two simultaneous, completely independent analyses. In proper phylogenetic analyses, one has to run several runs and more generations, so the posterior distribution is properly explored. Here, we run the MCMC with few runs and generations only to avoid taking too much of our humble cluster. The samplefreq is the frequency with which the MCMC samples will be written to file: set it 100. Considering that we run 50000 generations and sample every 100th, how many samples of the posterior will we have sampled by the end of the MCMC run?

mcmc ngen=100000 nruns=2 samplefreq=100

After the MCMC is finished, we need to summarize the posterior distribution: the sampled parameters and trees are respectively kept in the .p and .t files. For this, we need to run the sump and sumt commands, which summarize the continuous parameters and trees that have been visited during the MCMC. In order to summarise the MCMC, we need to set the burnin, which sets the number of initial generations that we throw away. These samples will, of course, not be used for inference, so it is important to make sure that we remove the burn-in iterations. We typically discard 10 to 20% of the initial iterates: as we have 1000 samples, we discard the first 200.

sump burnin=200
sumt burnin=200

Remember that MCMC samples are only representative of the posterior distribution if they are converging. Open the .p files using Tracer, plot the trace plots for several parameters and assess convergence: has the plateau phase being reached, and are the runs converging to the same regions? The ESS (effective sample size) value is also a good indicator of convergence; this value should be higher than 100 for all the parameters. If this does not happen, we need to run the MCMC for an additional number of generations.

Once these commands are run, the maximum a posterior tree is produced. You can open the resulting tree (.con.tre file) in FigTree. According to the Bayesian phylogeny, do we have evidence that the gastroenterologist deliberately infected the victim with HIV-infected blood from one of their patients? Support your conclusions using the obtained branch support values.