Hello,

After a quick glimpse in publications I can see that often people use all the methods available to scan for positive selection. I can see the difference between the counting methods and bayes, but I can't think of a reason why to use all the possible methods as opposed to only the most comprehensive one (like FUBAR). Maybe FUBAR and MEME (to account for episodic selection) are enough, and ML methods not needed?

I'm working with complete mitogenomes, on two highly divergent mitochondrial haplogroups (ND2: 6.6%, in birds is species level divergence) and its sister species.

Thanks!

Hi Hernan,

I agree that it is not necessary to run all methods you can get your hands on (especially since some are strictly inferior to others).

My workflow suggestion would be this (not that all these methods are ML based; the only thing that differs is how significance is called)

1). Run FUBAR (because it is very quick) to look for evidence of pervasive diversifying selection at SITES
2). Run MEME (if desired) to look for evidence of episodic diversifying selection at SITES (if all is well, all or most of the FUBAR results should be in this list is well, because pervasive selection is a particular case of episodic selection)
3). Run BranchSiteREL (if desired) to look for evidence of episodic diversifying selection along LINEAGES.

Sergei

Thanks heaps Sergei!

Hello Sergei,

I followed your recommendation, thanks! I was wondering if you can help me with another question.
Is there a reported comparison of the sensitivity of FUBAR vs. MEME. I got way more positive sites with MEME, is that expected?
I'm working with within lineages fixed aa replacements only. With MEME i get some positive results in terminal branches, usually only 1 individual per lineage. This are the episodic signal of selection, isn't it?

Thanks for your time!

Regards,

Hernan



Sergei wrote on Mar 31^st, 2014 at 12:30pm:

Hi Hernan, I agree that it is not necessary to run all methods you can get your hands on (especially since some are strictly inferior to others). My workflow suggestion would be this (not that all these methods are ML based; the only thing that differs is how significance is called) 1). Run FUBAR (because it is very quick) to look for evidence of pervasive diversifying selection at SITES 2). Run MEME (if desired) to look for evidence of episodic diversifying selection at SITES (if all is well, all or most of the FUBAR results should be in this list is well, because pervasive selection is a particular case of episodic selection) 3). Run BranchSiteREL (if desired) to look for evidence of episodic diversifying selection along LINEAGES. Sergei