In approaching the problem, a wide variety of physical models have been used, ranging from simple lattice models which only consider nearest neighbor interactions to highly complicated simulations involving complete quantum dynamics and the simulation of thousands of water molecules. These two approaches are on opposite ends of the spectrum of possible models that could be considered, and neither is likely to work, the former due to its over-simplification of the physical situation, and the latter due to the tremendous computational resources required.

One method is to approximate potential functions for all of the amino acids. These functions need to be accurately defined at up to 4-6 Angstroms distance, and after that, must exhibit asymptotic behavior. Strict energy minimization algorithms are quite computationally expensive, but retain an advantage of several orders of magnitude over complete dynamic simulations. The other inherent difficulty is that proteins are not guaranteed to be stable only in a minimum energy configuration, and that the local minima found by the algorithm need not be close to the absolute minimum. While dynamic simulations tend to be prohibitively expensive, it is possible to do a limited simulation that treats the peptide chain accurately, but makes some approximations for the side-chains.