Hi Boka,
This option in the MCMCtree control file specifies the values of the alpha and beta parameters of the gamma distribution that will be used as a prior on "sigma2". This parameter (sigma2,
σ^2) is the variance of the logarithm of the rate in the log-normal model of rate change. Large "sigma2" (for example, 0.2 or 0.5) means serious violations of the clock, and small values (like 0.01) mean that the clock almost holds. For shallow trees, the clock may not be very wrong, and you can specify a small prior with mean "sigma2 = 0.05"
in the control file, and hence use something like
sigma2_gamma 2 40 in the control file (i.e., mean_Gamma = mean_sigma2 = alpha / beta = 2 / 40 = 0.05). For deep phylogenies, the clock may be expected to be seriously violated and you can specify a large prior mean so "sigma2 = 0.1", such as
sigma2_gamma 2 10. If you have a prior knowledge about "sigma2" based on other studies, you should center the gamma prior around the value of "sigma2" you believe is more plausible by changing "alpha" and "beta" accordingly in the
sigma2_gamma option in the control file.
You can read more about the prior on the rates used in MCMCtree
in "Tutorial 3" here and on page 47 in the
PAML documentation here.
Hope this helps!
All the best,
Sandra