Hello,
I was wondering if it was possible to detect positive selection using only a few sequences from a larger dataset.  For example, I have longitudinally sampled sequences from HIV infected individuals and I would like to know at what timepoint positive selection (dN) is the strongest.  Is it possible to use just portions of a larger tree and dataset to perform this type of analysis? Or would I have to create a subtree and smaller dataset for each set of timepoints I am trying to compare?  In the case that I have to make separate datasets, would I need to screen for recombination for each dataset, or could I just use the breakpoint positions from the total dataset, given that there is more power to detect recombination with a greater number of sequences?  Thank you!

-Crystal

sounds to me like a job for GABranch - though if the alignment is predominantly within-patient variation you might want to think about non-dN/dS methods such as directional REL.
- a.