Exactness of the cluster variation method and factorization of the equilibrium probability for the Wako-Saito-Munoz-Eaton model of protein folding

I study the properties of the equilibrium probability distribution of a protein folding model originally introduced by Wako and Saito, and later reconsidered by Munoz and Eaton. The model is a one-dimensional model with binary variables and many-body, long-range interactions, which has been solved exactly through a mapping to a two-dimensional model of binary variables with local constraints. Here I show that the equilibrium probability of this two-dimensional model factors into the product of local cluster probabilities, each raised to a suitable exponent. The clusters involved are single sites, nearest-neighbour pairs and square plaquettes, and the exponents are the coefficients of the entropy expansion of the cluster variation method. As a consequence, the cluster variation method is exact for this model.