DESIGN OF A KNOWLEDGE-BASED FORCE-FIELD FOR OFF-LATTICE SIMULATIONS OF PROTEIN-STRUCTURE

Prediction of protein structure from amino-acid sequence still continu es to be an unsolved problem of theoretical molecular biology. One app roach to solve it is to construct an appropriate (free) energy functio n that recognizes the native structures of some selected proteins (who se native structures are known) as the ones distinctively lowest in (f ree) energy and then to carry out a search of the lowest-energy struct ure of a new protein. In order to reduce the complexity of the problem and the cost of energy evaluation, the so-called united-residue repre sentation of the polypeptide chain is often applied, in which each ami no-acid residue is represented by only a few interaction sites. Once t he global energy minimum of the simplified chain has been found, the a ll-atom structure can easily and reliably be constructed. The search o f the lowest-energy structure is usually carried out by means of Monte Carlo methods, though use of more efficient global-optimization metho ds, especially those of deformation of original energy surface is pote ntially promising. Monte Carlo search of the conformational space can be accelerated greatly, if the chain is superposed on a discrete latti ce (the on-lattice approach). On the other hand, the on-lattice approa ch prohibits the use of many efficient global-optimization methods, be cause they require both energy and its space derivatives. The on-latti ce methods in which the chain is embedded in the continuous 3D space a re, therefore, also worth developing. In this paper we summarize the w ork on the design and implementation of an off-lattice united-residue force field that is underway in our group, in cooperation with Profess or H.A. Scheraga of Cornell University, U.S.A.