Hi,

I have a few questions regarding the data output in GA-Branch.

1.  I often get models where dN/dS is inferred to be 1000.  This obviously seems really high--how do I interpret (is it some sort of upper limit)? 

2.  If the best fitting model shows branches with dN/dS greater than 1, but the model averaged support values are relatively low, I'm not sure exactly how to interpret.  Is this basically like if you have a phylogenetic tree that depicts certain relationships but the bootstrap support is low for branches on the tree (I know it's not the same statistical procedure, but would it be interpreted similarly)? 

3.  I am having issues redrawing some of the trees.  I've attached one for reference.  In this case, I actually gave a sequence for another population that is sister to MOJBAJA.  Datamonkey discarded it because they are the same.  So, if I want to redraw the tree to include that population, the green branch with 26% model averaged support would be the branch leading to the common ancestor of these two terminal branches?  In that case, the two terminal branches would have no values (which I guess makes sense because there is no change)?

4.  Looking at the same tree, MOJCAT/MOJMOJ/MOJBAJA form a clade and AZTUS/AZCHP form a separate clade.  I am really interested in the branches leading to the common ancestors of these clades. I get information on the branch leading to AZ but not for the branch leading to MOJ.  I assume this is because I would need an outgroup? 

Note: Unfortunately, I have been unable to upload a picture of the tree.  It is only 49 kb (jpeg), but it keeps saying it "exceeds maximum allowed upload-directory volume."  Anybody know what I am doing wrong.

Thanks for your help!

Hi Sergei,

Thanks for getting back to me.  Here are URLs relating to questions 1 and 2 above.

For question 1:
Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login

I assume dn/ds of 10000 is for a situation where all substitutions are nonsynymous?

For question 2:
Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login

Also, you mentioned that GA-Branch has been superseded by BS-REL.  We find with our dataset that GA-Branch seems to have more power to detect selection when you have a high dn/ds across the entire gene, while BS-REL seems to have more power to detect selection when a small number of individual codons along a branch have experienced multiple changes.  Either of these patterns is of significance to us, so we have been using both tests.  Does BS-REL really replace GA-Branch as they seem complementary (though of course I may misunderstand something)?  Thanks for your help!