Dear Sergei,
My question relates to the suitability of my data for usage with Datamonkey.

I have sequenced environmental samples for a specific gene. Upon phylogenetic tree construction, they cluster with either of two separate species. However while one species is represented by approximately 30 unique sequences (I am saying this based on the tree), the other cluster is represented by 6 sequences. The diversity within each cluster is very low (over 99% identity), while the identity between clusters is ~96%. Would it be wrong to use the whole data set for selection analyses, since one cluster is represented by many sequences? Should I analyze each cluster seperately? (I assume the low divergence would be a problem). Should I use a subset of sequences from the large cluster?


Thank you very much for your answer,

Best,

Mete

Hi Mete,

Uneven sampling could indeed bias your analysis: imagine that a site is under negative selection in the clade with 30 sequences, but is under positive selection in the clade with 6 sequences. The former signal will probably dominate and show the site as being under negative selection. That said, all site-by-site selection analyses assume that there is no variation in selection strength between lineages.

I would run your data through GA Branch to look for lineage dN/dS variation, and assuming there is not too much (or that it is not localized to some parts of your subtree), run the FEL analysis.

Sergei

Thanks a lot..