Dear Luli,

TestBranchDNDS.bf tests whether the branches you choose has different dN/dS ratios from the rest of the tree (all branches in the rest of the tree are assumed to have the same ratio). The null hypothesis is that all branches are the same, and the alternative is that the ones you selected a priori are different. One possibility for your results is that the entire tree is (on average) under strong positive selection, and letting your branches have their own rates does improve the fit much. Could you perhaps paste the output here, so that I could see if that is a good possibility? Another way to check is to run AnalyzeCodonData.bf analysis using a local codon model (i.e. estimating dN and dS separately for all branches) and seeing if most of the dN/dS ratios in that model are > 1.

One can indeed construct meaningful evolutionary tests for gene families using dN/dS (or codon in general) models. Take a look at Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login for some ideas on tests to be done in a similar setting.

Did you have anything in particular you wanted to test?

Cheers,
Sergei