-
Slides day 1
Exercise 1 - Parameter estimation
Exercise 2 - Tree topologies
Exercise 3 - Model comparison
Exercise 4 - Branch support
Exercise 5 - Command line
Exercise 6 - Inferring ML phylogenies with codon models
Exercise 7 - Inferring ML phylogenies using real datasets
Exercise 8 - Re-Analyze published datasets
Solution 2 - Tree topologies
Inferring phylogenies using maximum likelihood
Observing the effect of tree search routines on the initial inferred tree topology.
Goals
In this exercise you are asked to optimise the tree topology on the substitution parameters obtained using ML performing a tree search (i.e. NNI, SPR, TBR) on the initial tree topology.
Execution
1. Run
- Nucleotide substitution model = HKY85 + Gamma
- Estimating transition/transversion ratio ( parameter of HKY85 model)
- Estimating alpha parameter (remember for gamma distributions used in phylogenetics)
- Estimating nucleotide frequencies with ML
- No tree search (tree optimisation)
Here is the list of the parameters to change from the PhyML menu:
From 2nd menu
[M] ................. Model of nucleotide substitution HKY85
[F] ................. Optimise equilibrium frequencies yes
[T] .................... Ts/tv ratio (fixed/estimated) estimated
[C] ........... Number of substitution rate categories 4
[G] ............. Gamma distributed rates across sites yes
[A] ... Gamma distribution parameter (fixed/estimated) estimated
From 3rd menu
[O] ........................... Optimise tree topoLOGy no
Questions
1. Compare the trees obtained with and without tree-search. What do you observe and why?
The topology inferred without tree search presents a variation in the internal node attribution for the clades (Saki,Titi).
With tree-search | Without tree-search |
---|---|
2. Compare the model estimates with and without tree-search. What do you observe and why?
The transition/transversion ratio differs slightly between the two runs. As the matter of fact, parameter appears to have a higher value when tree-search is not performed.
3. Compare the likelihood of the ML and NJ trees. What do you observe and why?
The likelihood value obtained without tree-search is lower than the value obtain performing the topology optimisation. This value is expected since the substitution parameters are optimised over the first inferred tree topology which did not undergo any refinement.
with tree-search: -6172.58045
without tree-search: -6173.00555
phylogenies tree-estimation maximum-likelihood parameter-estimation
- Previous
- Next