| This work presents a pilot study of a novel algorithm for perturbative protein folding by free energy minimization. It combines potential energy smoothing using Gaussian convolution of the Boltzmann probability and hierarchical macrostate dissection of conformation space. These approaches are complementary. Smoothing deforms the potential energy function to reduce the complexity of the rugged protein potential energy landscape, but requires an efficient search of conformation space to identify regions of low free energy. Macrostate dissection identifies and tracks meta-stable macroscopic states of probability (macrostates) to efficiently explore conformation space, but requires a guide to navigate through the proliferation of macrostates to reach the global free energy minimum macrostate. The goals of combining these two methods are to allow fewer macrostates to be followed at each annealing step and improve the efficiency of stochastic sampling within each macrostate.; Towards these goals, a set of object-oriented, parallel processing programs were developed to implement potential smoothing and macrostate dissection. A novel smoothed form of the ECEPP/3 potential was derived by approximating the Gaussian Convolved Probability Smoothing. Challenges unique to macrostate dissection and smoothing, including tracking merging and shifting macrostates, were overcome. A new scheme for selecting which macrostates to follow at each annealing step, based on the minimized potential energy, was found to be more efficient for global optimization of the peptide Met-enkephalin than existing schemes based on thermodynamic properties of macrostates.; A key element to a successful combination of potential smoothing and macrostate dissection lies in the relationship between the smoothing and temperature annealing schedules. In principle, the combined method would allow them to be adaptively determined. However, to begin exploring the effects of smoothing on macrostate dissection, several simple annealing schedules were tried. An annealing schedule was discovered that reduced the number of macrostates required for global optimization of Met-enkephalin. This result and results suggesting increased macrostate sampling efficiency may be obtained using an adaptive cooling schedule bode well for future research integrating an adaptive annealing schedule into the smoothing and macrostate dissection methodology. |
