Biological macromolecules, large chain molecules with hundreds of torsion angles, adopt compact, uniquely folded and rigid conformations that correspond to their global free energy minimum. Predicting this unique conformation from a vast number of alternatives, for the whole protein or its parts, is the biggest challenge of computational biology. One of the difficulties is conceptual. To evaluate the free energy correctly we need to account for the dynamic nature of the entire system, including mobile water molecules, flexible side-chains and soft vibrational modes of a solute. Molecular Dynamics (MD, reviewed in Ref. 1-4) or Monte Carlo simulations (MC, reviewed in Ref. 4-8) in water can be applied to sample the conformational space and evaluate the free energy. However, these methods are still too slow to reach the biologically relevant folding times for proteins or even large peptides.