A parallel genetic algorithm for polypeptide three dimensional structure prediction. A transputer implementation

We propose a hybrid genetic algorithm to carry out a mapping of the conformational space of polypeptides, and at the same time we suggest a fitness function that incorporates quantitatively many factors that experimentally are known to influence the process of protein folding. Some of these factors are the hydrophobicity of the molecule, the disulfide bond among cysteine sulfur atoms and the compactness of the molecule. The steric energy of the molecule is calculated using molecular mechanics force fields. To account for the packing of the side chains of the polypeptide, the conventional GA is hybridized, endowing it with a local optimization function to perform the task. Moreover, due to the huge number of conformers the system processes, and because of the nature of the fitness function it manipulates, it has been parallelized, and its implementation on a network of transputers is also discussed.