Generalized Potential Function Smoothing Algorithms for Conformational
 Energy Optimization

 Reece K. Hart, Rohit V. Pappu and Jay W. Ponder

 Biophysical Journal, 74, A176 (1998)

 Abstract:

 Locating the thermodynamic ground state is of importance for systems such
 as clusters, proteins and glasses. The potential energy surfaces (PES) for
 these systems are sufficiently rugged to preclude the possibility that
 local optimization methods will find the global minimum from random
 starting conformations. This is frequently referred to as the multiple
 minimum problem. Local optimization methods are oblivious to the large
 scale features of the PES and typically become trapped in one of the
 overwhelming number of local minima. Two important paradigms have emerged
 for solving the global optimization problem: 1) Simulated Annealing and
 2) Potential Smoothing. In smoothing, the original rough PES is transformed
 through systematic application of a smoothing operator to one which may be
 variably deformed. This has the effect of gradually reducing the number and
 size of features of the surface to improve conformational smapling. We have
 derived computationally efficient and deformable versions of the standard
 AMBER/OPLS and MM2 force fields using the diffusion equation method
 introduced by Scheraga and coworkers. We present results which describe
 1) the structural determinants of the smoothing process; 2) the nature
 of the surface at various levels of deformation; 3) applications to Argon
 clusters, N-Acetyl-Ala-Ala-N-Methyl tripeptide, cycloheptadecane, and
 rigid body helix docking; and 4) deterministic and stochastic secondary
 search procedures by which basins on a smooth surface may be ascribed to
 sets of low-energy conformations on the original surface.