Knowledge-based potentials in protein fold recognition

An accurate potential function is essential for protein folding problem and structure prediction. Two different types of potential energy functions are currently in use. The first types are based on the law of physics and second type is referred to as statistical potentials or knowledge based potentials. In the latter type, the energy function is extracted from statistical analysis of experimental data of known protein structures. By increasing the amount of three dimensional protein structures, this approach is growing rapidly. There are various forms of knowledge based potentials depending on how statistics are calculated and how proteins are modeled. In this review we explain how the knowledge based potentials are extracted by using known protein structures and briefly compare many of the potentials in theory.