Use Of Phylogeny In Branch-Site-Specific Selection Model

Question

[Eric]

Hello,

I am investigating several genes using the Branch-site REL program on datamonkey. If my gene is, e.g., BRCA1, should I construct a phylogeny based on BRCA1 or should I use a true phylogeny (i.e., one which accurately reflects the relatedness of the species)?

Thanks.

NB. I have posted this question on the datamonkey message board, but no response has been forthcoming.

Answer 1

[DG]

Use an accurately constructed phylogenetic tree for the gene. Gene trees are different from species trees and for many, many reasons the real evolutionary history of a gene tree can differ from the overall species tree. Duplication and differential loss, horizontal gene transfer, incomplete lineage sorting, gene conversion, etc. When doing tests for selection you should always be inferring a phylogeny based on the gene/protein of interest and not using a reference species tree. Remember that branch lengths and other parameters are very, very important in the accuracy of your results so use a good Maximum-Likelihood or Bayesian phylogenetic algorithm with an appropriate evolutionary model.

[Eric]

Thank you. That's all very helpful.