For the positive selection here you cant predict an appropriate no of species it should be least two and more and more for the best results.but in mammals the closely related species are not too plenty so you should not bother for more species. for evolutionary distance you should take closely related species in distinct environments .

Hello Ozge,

For selection analyses you want to be in a range where your rate of synonymous substitutions (dS) is not saturated. If this rate is too high then any methods based on dN/dS will not be appropriate.

In mammals, Rodents or Simian primates are great sets of species to detect signatures of selection (over 20 assembled genome in each). In general I would say that between 50-70My of evolution is a good amount for selection analyses in mammals.

As far as number of species, there is a great study by Sarah Sawyer's group that looked into this. I would recommend to aim for 20. If you are using likelihood methods you will run into computational hurdles with more species. But I would advise against using the entirety of mammals to run any type of selection.

I hope it helps.

Cheers,

Antoine.