Torsion Angle Refinement and Dynamics as a Tool to Aid Crystallographic Structure Determination
Crystallographic methods using experimental diffraction data have produced about 85% of the macromolecular structures in the Protein Data Bank. Before deposition, nearly all crystal structures are refined with gradient-driven optimization techniques. Refinement is typically performed with iterative local optimization methods. A common problem is convergence to local minima. Reparameterization of the model in torsion angle space reduces the number of parameters. This in itself can help to escape from local minima. Combination with rigid-body dynamics algorithms results in an important tool for sampling conformational space. This paper presents the torsion angle refinement and dynamics algorithms implemented for the phenix.refine program and the results of various tests.