Predicting Protein Tertiary Structure using a Global Optimization Algorithm with Smoothing

We present a global optimization algorithm and demonstrate its effectiveness in solving the protein structure prediction problem for a 70 amino-acid helical protein, the A-chain of uteroglobin. This is a larger protein than solved previously by our global optimization method or most other optimization-based protein structure prediction methods. Our approach combines techniques that “smooth” the potential energy surface being minimized with methods that do a global search in selected subspaces of the problem in addition to locally minimizing in the full parameter space. Neural network predictions of secondary structure are used in the formation of initial structures.