CGU: An Algorithm for Molecular Structure Prediction

摘要: A global optimization method is presented for predicting the minimum energy structure of small protein-like molecules. This begins by collecting a large number molecular conformations, each obtained finding local potential function from random starting point. The information these conformera then used to form convex quadratic underestimating all known conformers. underestimator an L1 approximation minima, and linear programming formulation solution. this predict function, allowing localized conformer search be performed based on predicted minimum. new set conformers generated serves as basis another underestimation step in iterative algorithm. algorithm has been structures heteropolymers with many 48 residues, can applied variety models. results also show dependence native conformation sequence hydrophobic polar residues.