I'm afraid it's not that simple. Modified AA frequency will depend in a complex way on original AA frequency and number of allowed modifications per peptide. Each peptide can have different number of the given AA residues , so modifying say, 1 S per peptide will produce 50% modified S frequency only for peptides with exactly 2 serines.
More precisely, if P(l, n) is number of peptides in your proteome with length l and number of serine residues equal n then frequency of modified serine will be sum(P(l, n) * min(mods, m))/sum(P(l, n) * l) where summation goes over whole proteome.
Bill, please correct me if I'm wrong here.