Hybrid Genetic Algorithms for Minimization of a Polypeptide Specific Energy Model

A hybrid genetic algorithm for polypeptide structure prediction is proposed which incorporates e cient gradient based minimization directly in the tness evaluation Fitness is based on a polypeptide speci c po tential energy model The algorithm includes a replace ment frequency parameter which speci es the probability with which an individual is replaced by its minimized coun terpart Thus the algorithm can implement either Bald winian Lamarckian or probabilistically Lamarckian evolu tion Experiments are described which compare the e ective ness of the genetic algorithm with and without the local minimization operator and for various probabilities of re placement The experiments apply the techniques to the minimization of the ECEPP energy model for Met Enkephalin Using tness proportionate selection the hybrid ap proaches obtain better energies and better basins of at traction than the standard genetic algorithm and often nd the global minimum When tournament selection is used the results are qualitatively similar except that the hybrid approaches are prone to premature convergence