Search for native conformations of organic molecules by genetic algorithms

Recent results concerning the efficiency of genetic algorithms for structure prediction of organic molecules are presented. The main focus of this paper is on the efficiency of different problem specific crossover operators and repair heuristics. A fluorescent labeled nucleoside and an oligopeptide are used as application examples. The molecule structure adapted crossovers and repair heuristics turned out to increase the robustness of the genetic algorithm significantly. In combination with an island model they increased the probability of finding the global optimum of the nucleoside from 5% to 90%. Using a genetic algorithm with problem specific crossover we found a new global optimum for Met-Enkephaline and the ECEPP2 force field.