Codeml Pairwise Comparison
3
2
Entering edit mode
13.1 years ago
Juliana Soto ▴ 50

Dear all,

i´m analyzing two genomes (Genome 1 and Genome 2) and i have pair genes from those two related species. I want to know if the positive selection happened.

For this, i want to use the Branch test in codeml to test the significance of the analysis determining the maximum likelihood ratio with the following hypothesis:

Ho: one ratio (model 0, fix_omega 0, neutral)

H1: free ratio (model 1, Fix_omega 1, variation)

but i´m not sure if it is the appropriate method to identify genes that are statistically under positive selection in my pairwise comparison.

I know that for pairwise comparison i have to use model -2, but i don´t know how to test the hypothesis under this runmode value.

So the questions are : 1.Are the parameters for testing the hypothesis good? or How can I set them? 2.In order to validate my hypothesis using runmode -2, How can I set the parameters now in the control file?

for that reason, I would like to know your opinion about it.

thanks

best regards

codeml pairwise • 7.5k views
ADD COMMENT
0
Entering edit mode

do you have only 2 species? no outgroup? if yes, I do not think it would be feasible to test for positive selection in branch. You should retrieve at least 2 outgroups.

ADD REPLY
4
Entering edit mode
12.7 years ago
Andre Elias ▴ 110

Hi,

If you intend to check for positive selection first I'd advise to use an alignment with much more than 2 species and, also, the inclusion of an outgroup (as mentioned by fransua).

Then, as far as I remember, the parameter you want to edit in the codeml.ctl (the control file) is "NSsites". If you set it to "NSsites = 0 1 2" it will test for H0, H1 and H2.

Basically what you'll do to compare hierarchical models regarding their omega values. H1, "nearly neutral" allows your omega to be <1 and =1, while H2, "positive selection" will allow omega to be >1 (so it has three categories of omega instead of two). This comparison is tested to check if these new parameters increase significantly the maximum likelihood values, and in order to do that you must perform a likelihood ratio test (LRT). It's very straightforward, being 2*(lnL1 - lnL2) = chi^2 (the number of degrees of freedom is the number of aditional free parameters).

(I hope this long text wasn't too boring or trivial)

If you're interested in positive selection analyses, I'd recommend you to take a look the HyPhy package, available here: http://www.datamonkey.org/ - my experience is that their results are not too different from one another, but never 100% the same. And their website has a lot of good information and references. :) I think it's a good idea to use both methods. Oh, if recombination is something you worry about, you should definitely check their PARRIS algorithm. :-)

ADD COMMENT
1
Entering edit mode
13.0 years ago
Juliana Soto ▴ 50

thanks a lot for your answer!! now i know how to run the analysis

best

ADD COMMENT
0
Entering edit mode
13.0 years ago
Juliana Soto ▴ 50

other question:

I have more than two species but i want to do a pairwise comparison between strains using 2 genomes. for that reason i´m using codeml runmode –2 and i don´t know if for this analysis i need a outgroup because i´m using model =0.

I´m confusing about it because i don´t know how to define the parameters in order to prove the hypothesis (null vs alternative). I used this:

Ho: model 0, fix_omega 1, neutral

H1: model 0, Fix_omega 0, variation

are these parameters appropriate? or How can I set the parameters in the control file using runmode= -2?

thanks a lot

ADD COMMENT

Login before adding your answer.

Traffic: 1416 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6