We are currently working on a study of an immune gene locus in a complex of recently diverged, closely related vertebrate species. To test for selection on the sequences we recovered, we pooled alleles from all species together, created a phylogeny in MrBayes, and tested for selection using MEME and FUBAR. However, a colleague raised the concern that these methods might only be appropriate for variation that is fixed for each species (some of our species share alleles). Is this an accurate claim? Is iFEL a more appropriate tool, even though these are technically species, not populations? Any clarification on this issue would be greatly appreciated.

Hi there,

1). Generally, MEME will be much more powerful than iFEL: for example the latter will only detect selection if ON AVERAGE omega > 1 on all internal branches. Hence you are likely to miss episodic selection.

2). It is possible that in some of the species, selection is not fully resolved; however since in this case they are starting with the same allele, it is likely that you are simply not picking up some adaptive change because it has not yet occurred.

3). iFEL was originally designed to deal with situations where there is a lot of variation in terminal lineages (e.g. individual viral isolates), which has not been passed through the selection filter (e.g. due to onward viral transmission). In that case, it may be possible to mistake neutral variation (or something beneficial only to a single host, but deleterious on average, e.g. a rare CTL allele escape) for selection.

My advice would be to just run MEME.

Sergei

Thanks Sergei!
So, it's unlikely that in a relatively long-lived vertebrate species, slightly deleterious non-synonymous mutations segregating in closely related species would somehow be combined with mutations under positive/diversifying selection to produce some false positives? That's our concern about using the most sensitive methods like MEME.