Hello. Thanks for your help!
I'm using codeml to test patterns of selection among different species with a group genes. I have noticed in many research paper that genes need to be filtered with dn/ds>3 or ds<0.01 which will make the analysis inaccurate. In my case, I have noticed this problem for large proportion of the genes, especially large ds values, indicating saturation. I think this is due to the high divergence of the genes. However, I have chosen already the conservative genes and trimmed it carefully.
I used trimal with option -automated1, I want to ask if this is proper and can I remove high divergent sites in order to get this analysis done or my genes are actually not eligible for dn/ds analysis? I also tried with removing the divergent lineages.
Another concern is some genes works fine with site models but saturated ds values with branch models. Can I not care about the branch model results if the branch model is less fit for the data?
I appreciate your help!