HYPHY
implements the method of Felsenstein and Churchill (1996) (
as described in "A Hidden Markov Model Approach to Variation Among
Sites in Rate of Evolution". Mol. Biol. Evol. 13(1) :93-104),
with the extension that the rates are not fixed a-priori, but rather
sampled from one of the predefined distributions,
and the auto-correlation parameter lambda is estimated.
The Hidden Markov chain model assumes
all rates to be equiprobable and the uses the transition matrix with
entries as follows:
where N is the number of rate categories, and lambda is in [0,1].
If we define
then L+1 is the expected length of a patch of sites with the same
rate.
|