Datamonkey vs. Codeml - differences in results

Question

[nichlieh]

Hi all,

I am currently running positive selection tests in PAML and decided to run my data through Datamonkey as well just to get an idea of positive selection. I expected the results would be different as they are different methods. Even though alignments seem find and no errors were encountered the differences in positive selection are really huge. I was wondering if anyone could give me insight as to how these two programs are both meant to be useful platforms while yielding completely different results?

[nichlieh]

Sorry also, forgot to mention I am using GA-branch model!

Answer 1

[Brice Sarver]

Datamonkey is not really an approach in and of itself; it's a front-end to a cluster that allows you to perform many different analyses.

One of the benefits of the server is that it allows you to search for positive selection under many different models/approaches (FEL/IFEL/FUBAR/etc.) and then compare the results. Some approaches, like SLAC, are quite conservative - others are more liberal. The PAML manual describes the models implemented in codeml, and some of the other approaches on Datamonkey are described in two papers by Sergei Kosakovsky Pond et al. in 2005 (Bioinformatics and MBE); FUBAR is described in a paper by Ben Murrell et al. in 2013 in MBE.

Remember, these are different models and approaches for identifying sites under positive selection. It is not necessarily the case that they'll return exactly the same results. What might be more interesting is when they agree. If they disagree, perhaps you can explain the discordance by appealing to differences among models.

Answer 2

[nichlieh]

That makes sense, that you!