How To Set Tim1 Model In Mrbayes

Question

[Adrian Pelin]

Hello,

jModelTest suggested using the TIM1 model for my analysis. I got this as output:

Model selected: 
   Model = TIM1+I+G
   partition = 012230
   -lnL = 29285.3110
   K = 90
   freqA = 0.2754 
   freqC = 0.1487 
   freqG = 0.2840 
   freqT = 0.2919 
   R(a) [AC] =  1.0000
   R(b) [AG] =  2.3875
   R(c) [AT] =  1.1204
   R(d) [CG] =  1.1204
   R(e) [CT] =  3.6600
   R(f) [GT] =  1.0000
   p-inv = 0.1020
   gamma shape = 0.6270

Any idea how to set this up in MrBayes? Adrian

Answer 1

[David W]

I'm pretty sure MrBayes only let's you set 1 paramater (Jukes Cantor or F81), 2-paramater (Kimura or HKY) or 6-paramater (GTR) model.

I'd check out the AIC differences between the HKY models and GTR models in jModelTest and decide if steping up to the the 6-paramaters model is justified.

[Adrian Pelin]

delta AICc between HKY and GTR is 31, and if it is more than 10 it is significant, correct?

Yes, that is what I understood as well after extensive reading, that it does not allow setting a 4 parameter model. What I ended up doing in the end is running my analysis with a Mixed model, which is supposed to try out every possible model, including JC, GTR, etc. I already ran the analysis with GTR, and it looks okay, just want to make sure that the values do not change for branch support when I do the Mixed model.

[David W]

What you've done sounds very sensible to me (and yes, 10 is often quoted as a rule-of-thumb for a substantial difference in model-fit)