codeml for highly conserved sequences

[José Enrique Mejía]

Dear All,

This is my first post to the group. I am running codeml over a codon alignment for > 40 species of mammals to compute sitewise omega. I expect most of the sitewise omega values to cluster close to zero, seeing as the values for pairwise omega are within a narrow interval around 0.07.  To my untrained eye, however, none of the usual models in codeml (M0-M3, M7, M8) seem geared to testing for or against strong purifying selection specifically, as seems to be needed here. What would be the best practice for this? Are there any models more appropriate than the above? Thanks a lot in advance for your thoughts.

All the best,

Jose

[Ziheng]

you can run M0 either with and without constraining omega to be 1.

run the null hypothesis with

fix_omega = 1

omega = 1

then the alternative hypothesis with

fix_omega = 1

Compare twice the log likelihood difference with chi square with df = 1 (significance value 6.63 at 1%).  presumably you will reject the null in virtually every dataset.

every functional protein is under purifying selection, so the test will not excite people.

ziheng