Abstract
A high directional Monte Carlo procedure that predicts the topology of the energy hyper‐surface before it walks, is developed as a method to obtain a global energy minimum structure of polypeptides and proteins. It calculates its covariance matrix, which controls the individual trial step distribution of the next set of steps, from the second moment of the actual walk segment in the previous set. The method is successfully applied to the pentapeptide Metenkephalin system. And it is shown that some initial heating process, which provides the more flexible molecule, is necessary in order to overcome the energy barriers that can be overestimated by some biases in the empirical description of the system. The sampling efficiency, traced by an average conformational changes, is found to be at least 20 times greater than the one in the conventional Metropolis Monte Carlo methods, and it is expected that this increases in efficiency will be more prominent when the system is larger.