I would use the ModelTest to observe what would be the model of nucleotide substitution for my datasets, and I noticed he left a version with GUI and now more user-friendly, well I ran the test and the result came out with the AIC. But the model suggests that it does not match those of the HYPHY, I say this because, for example when choosing rate parameters, the user option, the HYPHY already placed on the AG, fixing him in 1.0000, when for example the result of Jmodeltest the value 1.0000 is AT = GT, as follows:

Model selected:
  Model = TPM3uf+G
  partition = 012012
  -lnL = 10499.5544
  K = 154
  freqA = 0.3723
  freqC = 0.1491
  freqG = 0.1598
  freqT = 0.3188
  R(a) [AC] =  1.6489
  R(b) [AG] =  13.1093
  R(c) [AT] =  1.0000
  R(d) [CG] =  1.6489
  R(e) [CT] =  13.1093
  R(f) [GT] =  1.0000
  gamma shape = 0.3080



My question is: How can I use the JModeltest results to set up HYPHY?

Best Regards,

Jackson Lima

Hi Jackson,

You HAVE to fix one of the rates at 1 (JModelTest seems to be doing it with GT or AT); you can always rescale everything relative to AG, e.g. in your case set AC = 1.6480/13.1093, AG = 1, AT = 1/13.1093 etc.

Generally speaking, I would just recommend that you use GTR; there is very little computational cost to adding more parameters, and all you may lose is a bit of precision.

What do you want to use this model for in HyPhy?

Sergei