Rev Language Reference

Constructs a rate matrix from 61 scaled selection coefficients w[i] and a 4x4 nucleotide mutation rate matrix mu(i,j). In the original paper the nucleotide mutation rate matrix is a GTR rate matrix. The FMutSel0 model differs from FMutSel by constraining all codons for the same amino acid to have the same scaled selection coefficient. The function fnMutSel differs from fnFMutSel by taking a codon mutation rate matrix. A substitution from allele i -> j can be decomposed into (1) all individuals initially have state i (2) a single individual mutates from i -> j, at rate mu(i,j) (3) the allele j goes to fixation Then the substitution rate Q is then given by Q(i,j) = mu(i,j) * Pr(j goes to fixation | i was fixed previously). The probability of fixation is determined by scaled selection coefficients: F[i] = 2*N*s[i] and the initial frequency 1/N of allele j.

er ~ dnDirichlet( v(1,1,1,1,1,1) )
nuc_pi ~ dnDirichlet( rep(2.0, 4) )
F ~ dnIID(61, dnNormal(0,1))
omega ~ dnUniform(0,1)
# The FMutSel model from Yang and Nielsen (2008)
Q1 := fnFMutSel(fnGTR(er, nuc_pi), F, omega)

# The same -- fMutSel = GTR(er,nuc_pi) + X3 + MutSel(F) + dNdS(omega)
Q2 := fndNdS(fnMutSel(F, fnX3(fnGTR(er, nuc_pi))), omega)