Can I set up (or define) more than one foreground branch in branch-site model A?

[Feifei Zhang]

Hello everyone,

When I am using branch-site models, with prior knowledge, I expect positive selection could occur on more than one branch in the tree.
It seems I have two options:
Option 1: conduct multiple tests, in each test, I set up one of the interested branches as foreground branch in model A; 
Option 2: only conduct one test, in this test, set up all interested branches as foreground branches in model A.

Which option is reasonable (or correct)?

Thank you,
Feifei


 

[Ziheng]

Both are possible, but it depends on your biological hypothesis.
ziheng yang

[bru...@fieldmuseum.org]

I would like to follow up on this question as I have a similar situation. I am interested in testing for convergent selection in multiple branches. While it is possible that the same sites are under selection in each of these branches, it is not required for the gene to be of interest, as long as each of the target branches is under some sort of positive selection. Therefore, I think that I should conduct multiple tests, one for each target branch and excluding the other target branches. Conducting one test seems like it would require the assumption that the same sites are under selection in each target branch.

Does this sound reasonable?

Thanks,

Ben

[Ziheng]

Ben's comments about the differences between the two options (two models) are entirely correct. It is then a biological question which option you would like to use.
If none of the models fits the biological truth, you can do the analysis in both ways, and then try to interpret the results, bearing in mind the assumptions made in the two models.
ziheng

[hvk]

Just a follow up to this bit of Ben's comments "

... one for each target branch and excluding the other target branches"

By "excluding", would you mean "to remove physically" the other target branches from the analysis?? Or, they remain in the data-set but not labeled?

Thanks,

Hossein

[bru...@fieldmuseum.org]

What I meant was to actually remove those branches from the analysis. Not just removing their label but excluding them completely.

Ben

[Ziheng]

If the branch is in the tree given your data, you have to decide whether it should be foreground or background.
No assumption, no inference.
Ziheng

[ymw...@gmail.com]

I have relevant question but regarding the clade ModelC.  If I want to test convergent seletion on several species that are, of course, not in the same clade, can I assign the branches leading to these species with the same branch nubmer (i.e., #1) and run the tree for the "clade" ModelC?  So, the question is whether the clade Model C require real clades in the tree, or random branches combined are also fine?

Thanks,

Chih-Ming
 

[cajawe]

Clade model C does not require real clades.