Hi

1) You can calculate the degree of freedom by subtracting the number of parameters (np) of H0 codeml output file (model=0 in this case) from H1 codeml output file (model=1 in this case). Number of parameters is specified by np variable in the output file.

2) In this case I would do the codeml Branch-site model, it will out the branch and sites under selection. And will pick only those genes which are having at least one site with BEB confidence >95%. This test will let you know which species gene is under selection pressure along with the sites under selection. In further analysis you can check if the selected sites are present in the functional domain of the protein, by visualizing its structure using PDB database (If its structure has been reported).

3) In case of Branch-site model one has to make ctr file like this:

For H0: model=2, NSsite=2, fix_omege=1, omega=1.

For H1: model=2, NSsite=2, fix_omege=0, omega=1.

2∆l=lnL1-lnL0

d.f = np1-np0

I would do LRT, BC and FDR at 5% and 1% level of significance, and consider the sites >95% BEB confidence with FDR 1% level of significance.

Hope this will be helpful,

Best wishes, Rahul

Hi

Finding positively selected sites with great accuracy is very tricky. To find positively selected sites within certain lineage, you have to mark that lineage as foreground in the tree by labeling it with "#1" or "$1". Selected genes can be predicted using LRT test and multiple hypothesis testing using BC or FDR @ 5 or 1% level of significance. Further positively selected sites you can pick from the output file, if the BEB confidence is >95% or 99%, will depend on your choice.

Best, Rahul

Thanks a lot, Rahul! If we were to see if there is positive selection among a certain branch, we conduct test for cases where we choose model = 0 (fixed rate of change for all branches) and choose model = 1(allow the branch that we specified to change freely). We can get the likelihood ratio for both tests, and conduct a chi-square test to see if there is significant difference between the w of our branch vs. the background w. I wonder what the degree of freedom is, how we can estimate that.

The test that involves choosing two models (model =0 vs. model =1) makes sense and allows us to determine if there is a significant difference between our foreground and background. Also it gives us w of our branch of interest and background branch. If when model =1 (free model), we found that our branch w has higher value compared to the background w(when model =0, i.e. all branches have the same w), what can we conclude? Can we say that our branch has relaxed evolution of rate compared to the background branch? Or our branch has faster evolutionary rate?

However, when examining if there is sites that evolve under positive selection, the parameters that people set makes me puzzled. They use branch-site model, so they set model =0, NSsites = 0 1 2. The output result shows that all branches have the same omega, why is this? I know that NSsites = 0, 1, 2 are testing cases of neutral evolution, positive and negative selection. Is it because they are allowing sites to be changed, so the whole branch keeps fixed? And that's the model that they detect if there is positive selected sites?

I run this in paml where I want to test if there is some site that undergo positive selection for my branch(which contains about 15 seqs and I labeled this lineage as #1), I tested no sites that are significant under 5% or 1% significance level. And the result gives me the same w for all the branches. At that time, it puzzled me. Now if my understanding is right, that makes sense right? Anyone can confirm if I am right? (My understanding is that when testing sites of positive selection among a lineage, model =0, assuming all branches have the same w, but allowing different sites to evolve differently?)

Then it comes to me another question, will there be a difference if I set the background of my tree differently ? Will it gives me different results? When running paml, what is the criteria of choosing background taxa or seqs if we want to get significant results?

Thank all for your attention! Jenny