Hi all,

I have sequence data for a total of eight taxa and would like to test for positive (or, failing that, relaxed) selection in three of these taxa, none of which are each other's closest relatives. From what I have gathered, eight taxa is not enough to get dependable results from a BranchSiteREL test and, in these situations, the GA-Branch test may be better. However, this latter test seems like it has been pretty aggressively phased out in favor of BranchSiteREL. Is it reasonable to use BranchSiteREL with eight taxa and, if not, can you suggest a more appropriate test? Is the GA-Branch test still implemented?

Thanks,
Ben

Hi Ben,

The number of taxa are not what determines the power of BranchSiteREL -- sequence divergence and sequence length do. You are probably thinking of MEME (which IS powered by the number of taxa). You should be able to use BranchSiteREL without any issues. If you use the standalone version of BranchSiteREL in HyPhy you can also restrict the set of branches to test to only the three you had selected a priori. This will reduce the number of tests done and boost the power as well (less stringent multiple testing correction).

Sergei

PS We are also developing a test for relaxation of selection, based on the same formalism as BranchSiteREL, but using a different set of models. This test has not been published yet, but if you are interested e-mail Joel Wertheim in my group for details.

Hi Sergei,

Yes, I must have misread something. Thank you very much for your response. I definitely will contact Joel Wertheim to ask about the test for relaxed selection.

Thanks again,
Ben