Codeml Output Dn/Ds On Each Branch

Question

[GR]

Hi Everyone,

I need few suggestions for codeml analysis.

I ran my gene set on codeml and found positive selection in a subset of genes. From the positively selected subset, I am discarding the genes that have unusually low ds. In that case should I just check the ds of the gene of interest or the whole branch? The other question is regarding the codeml output. In the tree below follow the link for an image of sample tree. My gene of interest is at node number 9. I understand that dn/ds is the results of comparing 2 or more than two species then what is the difference between 73..8 and 73..9? Basically the ratios should be same here because we are just comparing two species but it is not the same in case (see below the codeml results). Can someone help me with these basic questions.

codeml output:

branch          t       N       S   dN/dS      dN      dS  N*dN  S*dS

68..69      0.185   1551.6    641.4   0.7845   0.0570   0.0727   88.5   46.6

69..6       0.650   1551.6    641.4   0.7845   0.2005   0.2556  311.1  163.9

69..70      0.123   1551.6    641.4   0.7845   0.0380   0.0485   59.0   31.1

70..71      0.309   1551.6    641.4   0.7845   0.0952   0.1214  147.8   77.9

71..5       0.705   1551.6    641.4   0.7845   0.2175   0.2773  337.5  177.8

71..4       0.367   1551.6    641.4   0.7845   0.1132   0.1443  175.6   92.6

70..72      0.238   1551.6    641.4   0.7845   0.0734   0.0935  113.8   60.0

72..73      0.351   1551.6    641.4   0.7845   0.1084   0.1382  168.2   88.6

73..8       0.431   1551.6    641.4   0.7845   0.1329   0.1695  206.3  108.7

73..9       0.799   1551.6    641.4   0.7845   0.2466   0.3144  382.7  201.6

72..7       0.455   1551.6    641.4   0.7845   0.1403   0.1788  217.7  114.7

68..3       0.914   1551.6    641.4   0.7845   0.2821   0.3596  437.7  230.6

Answer 1

[David W]

I ran my gene set on codeml and found positive selection in a subset of genes. From the positively selected subset, I am discarding the genes that have unusually low ds. In that case should I just check the ds of the gene of interest or the whole branch?

Without knowing why you'd do this, it's hard to know how you should do it.

My gene of interest is at node number 9. I understand that dn/ds is the results of comparing 2 or more than two species then what is the difference between 73..8 and 73..9? Basically the ratios should be same here because we are just comparing two species but it is not the same in case

Your dNdS ratios for these branches _are_ the same - but that's presumably because you've estimated a model with only one omega for the whole tree (model=0 in the control file?). If you had used a model with different values of omega on differnet edges, it's quite possible that the one leading from node 73 to tip 9 would be different than the one connecting the same node to tip 8. That's because tree-based dNdS methods take into account the whole evolutionary history of the sequences, not just the pair-wise comparions. (In this case, knowing that 9 and 8 also share a recnt common ancestor with 7 is important information.

I'm not entriely sure what you are trying to get out of your analyses - but it's probably a good idea to read up the PAML documentation so you can get an idea of which models correspond to the questions you want to ask.