Differential evolution for protein crystallographic optimizations

Abstract

Genetic algorithms are powerful optimizers that have been underutilized in protein crystallography, given that many crystallographic problems have characteristics that would benefit from these algorithms: non-linearity, interdependent parameters and a complex function landscape. These functions have been implemented for real-space optimizations in a new fitting program, MIfit, for real-space refinement of protein models and heavy-atom searches. Some programming tips and examples will be presented here to aid others who might want to use genetic algorithms in their own work.