Dear All,
In CODEML, I am facing a problem with log likelihood values. I get exact values for M1 and M2 site models while nearly same values for M7 and M8 . I tried changing the parameters such as omega initial value, kappa initial value, number of Gamma categories etc. Also, I tried using the codeml.ctl file that is given in HIV NSsites examples. But all such attempts are futile. I get omega values and positively selected sites in result file. Only lnL values are causing problem.
Can anyone please suggest me a way out for this?
Identical lnL values for M1 and M2 (or M7 and M8) means that there's no evidence of "adaptive" codon substitutions in a given sequence using the site test. The way to test whether there's evidence for positive selection using PAML is to use the likelihood ratio test (LRT), which is calculated by multiplying the difference in lnL values by 2, and the p-value is extracted from the chi square distribution with 1 degree of freedom. In your case, the two lnL values are identical. Therefore, the LRT value—and thus, the chi square value—will be very low, and so the p-value will be very high. The "branch-site" test provides a bit more power, assuming you have a good number of sequences and substantial, but not saturated divergence between your samples.
- Badges
- Science method
- Similar topics
- Biological Science
- Evolutionary Biology
- Phylogenetics
- Phylogenetic Analysis