Paml - Branch-Site Models

Question

[Kartik Sunagar]

Dear All,

Sorry for the novice question. I have gone through the PAML documentation but it is still very confusing for me.

I have two lineages in my phylogenetic tree. I suspect positive selection in both these lineages, one of them being under a greater influence of positive selection than the other. I have run site models which prove this. Now, I want to present a phylogenetic tree specifying separate omega values for every branch that form these two clades and find out which are the species that are contributing the most to the over all omega value of the lineage.

To achieve this, I used the branch-site model (model=2 NSsite=0). But it prints out a tree where all the foreground and background branches have a single omega. So, what I want to know is ...

1. Is there any model that prints a tree with separate omega value for each branch in the tree (like the GA Branch test on the datamonkey server, which I can not use as it has a limitation of 20 sequences)?
2. Is this done using using the branch site model A (model=2 and NSsite=2) or the clade models (C and D)? I did try the branch site model A but it only prints out the probabilities for each amino-acid site to be under a site class and not the individual dn/ds ratio for each branch I am interested in.

I would really appreciate your help with this.

Thank you,
Kartik

Answer 1

[2184687-1231-83-]

Last time I checked, PAML didn't have an equivalent of Datamonkey's GA Branch model. You could download HyPhy and run the GA Branch model locally on your computer, the batch file is called BatchFiles/BranchSelector.bf

[2184687-1231-83-]

Yes, the models explained in the Lysozyme papers and the GA Branch method in HyPhy are different. GA Branch will create as many different dN/dS types as needed to fit the data, whereas as far as I know, in PAML you couldn't do that. In PAML you can do foreground/background labeling of branches, a Free-Ratios model where each branch can take a different dN/dS value, etc.

[Kartik Sunagar]

Thank you very much for the reply. I thought Prof Yang used a similar model in his paper "Likelihood Ratio Tests for Detecting Positive Selection and Application to Primate Lysozyme Evoltuion", figure 2. Is it a different test? Any idea which model will give me a similar dn ds values for each branch? Thank you very much!

Cheers,
Kartik

Answer 2

[bagdevi.mishra]

To get a tree with separate omega value for each branch in the tree as the output, one can use branch-based free-ratio model in codeml with the following paraemters.

model = 1
NS sites = 0