The protein folding problem is the problem of predicting the 3D structure of a protein when the linear sequence of aminoacids identifying it is known. In this paper we present a declarative implementation in Constrain & Generate style in CLP(ℱFD) of the protein folding problem, for models based on face-centered cubes. We use information concerning secondary structure (and other heuristics) to sensibly prune the search space. Preliminary results on real proteins are encouraging.