Modified crossover operators for protein folding simulation with genetic algorithms

In this thesis we cover the application of genetic algorithms to the protein folding problem, which is a very important problem in biochemistry with applications to medicine and drug development. We present two new ideas for improved crossover operators for the folding problem which use secondary structure information from the running population to weight the choice of crossover points. The operators were implemented and tested with a pre-existing protein folding simulator on a number of sequences and the resulting protein structures were compared to the experimentally determined structures for those sequences. They produced structures with 14% improvement in structural similary for one operator and 16% improvement for the other on average.