Codeml- Dn/Ds High Due To The Low Ds

Question

[GR]

Hi All,

I am new to codeml analysis. I have done the codeml analysis on a set of genes using both the lineage specific and site specific models and found a subset of genes under positive selection. Somebody has pointed out that the dn/ds of your genes could be high just because of low ds and this can not be considered as episode of positive selection. My question is does codeml models already control this and report only the genes that have higher dn/ds excluding the ones that dn/ds high due to not low ds?

Thanks, RT

Answer 1

[Biojl]

Codeml doesn't control for that.

We usually discard genes with branches showing dS < 0.01, as such low dS values may result in inaccurate dN/dS estimates, and branches showing dS or dN > 2 indicating saturation of substitutions.

You might also want to discard genes with abnormally high dN/dS (>10).

[GR]

Thanks a lot for reply. What causes the abnormally high dn/ds? Does this not represent the positive selection?

Do you write your own script to do this or there is any utility already available to calculate dn and ds out of codeml.

[Biojl]

Usually it's due to errors in the multiple alignment. We recently published a paper covering some of your questions. http://www.ncbi.nlm.nih.gov/pubmed/23377868

[GR]

Great! Thanks a million.

[Biojl]

I don't want to sound selfish, It's just a general comment. If you're happy with the answers you get you should vote up people answers.

[GR]

Thanks for help Biojl. In your paper you have mentioned that you calculated dn and ds using free ratio model in codeml. I have the output of free ratio model. But there is ds calculated for each branch of the tree under different w parameters. I am not sure how to extract ds from this output. I have posted this question as a separate post Need help with understanding codeml output. Can you please help.

Thanks for the great help. RT