Parallel genetic algorithms on PARAM for conformation of biopolymers

A software is developed using genetic algorithms to predict the structure of a polypeptide chain. The algorithm is based on the principle of evolution and it improves the solution of the posed problem by genetic operations crossovers and mutations. Dihedral angles (/spl phi/,/spl psi/) are taken as the basic variables for the structure of the molecules and genetic operations are carried over on a population of binary strings of (/spl phi/,/spl psi/) angles. First, a sequential code is developed in FORTRAN on a standard workstation. A parallel version of the program is implemented on a distributed computing platform PARAM, developed by CDAC. The methodology and the practical aspects of the algorithm is presented with case studies of a dipeptide and an octapeptide. The usefulness of the migration model, developed for the first time, in achieving efficiency is stressed. The migration model proved to be more efficient and the minimisation for the octapeptide improved from 2% to 10%. This improvement is expected to be more pronounced for larger molecules.