This project is a collaboration between members of the global optimization computer science and the computational biology communities to develop new methods for protein structure prediction and ligand docking. If successful, the methods developed here will help solve two important problems in computational biology - the need for fast conformational searching to predict the structures of proteins or complexes of proteins with other proteins and ligands, and the need to improve folding and docking models. Both problems are fundamental to understanding how the molecules of life work. This project will not solve these problems itself, but will develop new computational methods that can contribute to their solution.
Technically, there are two specific aims: (1) To develop efficient methods for searching conformational spaces to find globally optimal (native) conformations on energy landscapes. (2) To develop efficient methods for searching parameter spaces to find optimal parameters for the large, complex models that are common in computational biology. This work is based on Underestimator methods that do not search over the tops of energy landscapes like Monte Carlo, Simulated Annealing, and Molecular Dynamics, the current standard methods.
