Substitution model selection in BEAST

Question

[wes]

I have constructed phylogenomic tree using the protein-coding genes of plant mitogenome using maximum likelihood method. Now I want to compare it using Bayesian method using BEAST.

The alignment file is amino acids instead of DNA.

1. What is the best recommendation of substitution model to choose? The default is Blosum62
2. Can I left other setting such as Trees, Operators, MCMC as default?

Answer 1

[Mensur Dlakic]

IQ-TREE's ModelFinder will help you find the substitution model that works best. There is also a paper if you wish to read the details.

[wes]

Thanks Mensur Dlakic

The best-fit model according to AICc using ModelFinder is : HIVw+F+I+G4

However, may I know which substitution model should I choose as both BEAST version did not have HIVw option as shown in the printscreen below?

BEASTv.1.10.4

BEASTv.2.6.7

In addition, how to specify 'F' (empirical frequencies)? I only see empirical frequencies option if the HKY substitution model is chosen (nucleotide). However, my data type is amino acids.

[Mensur Dlakic]

I don't use BEAST, so you may get better advice from someone else in that regard. It is quite possible that it doesn't have the HIVw matrix, in which case I suggest you go down the list of your best models until you find one that is implemented in BEAST. If you can't specify the fixed frequencies, the program will treat them probabilistically and will likely still find the best solution with maybe some extra work.

[Biodeer]

https://beast.community/custom_substitution_models https://groups.google.com/g/beast-users/c/N1AQVTKgO6U?pli=1