The complexity of proteins' tertiary structures requires algorithms to efficiently explore the conformational space. BCL::Fold efficiently enumerates and scores possible conformations by placing predicted secondary structure elements (SSEs) in the three-dimensional space. The degrees of freedom of amino acids within SSEs are restrained by hydrogen bonds resulting in a small range of allowed dihedral angles reducing the complexity of sampling space. Sampled conformations are scored using knowledge-based potentials to approximate the free energy difference between to sampled conformations. For selected models loops and side-chains are added using a cyclic coordinate descent algorithm and rotamer libraries. The stability of selected models is evaluated using molecular dynamics simulations.