Finding energetically stable conformations of proteins is one of the most interesting and difficult problems in biology. Given the chemical composition of a molecule, its three-dimensional conformation is of interest because it is directly related to the function it is able to perform. Since some years, we have been working on an easier problem, where the identification of protein conformations is aided by information obtained by experiments of Nuclear Magnetic Resonance (NMR). Under certain assumptions, we are able to discretize the problem and use an efficient Branch & Prune (BP) algorithm for its solution. In this work, we move towards the more difficult situation in which the information given by NMR is not available. The original BP does not work in this case, because there is no information for performing its pruning phase, that is the strong point of the algorithm. In this paper, we study and present a new energy-based pruning device to be added to the BP algorithm. First computational experiences on a set of small homopolymers show that this approach is promising for the identification of low-energy conformations of molecules.