Conditional dynamics: refinement and phasing byN-particle optimization
Piet Gros, Sjors H. W. Scheres · Acta Crystallographica Section A Foundations of Crystallography · 2002
Exploiting the prior knowledge of standard protein geometry in structure refinement is common practice.The prior geometrical data is a wealthy source of phase information.However, the application of these data requires assignment of the protein structure.This is typically achieved by model building in an experimentally phased map.We have developed an N-particle formalism that allows a rigorous treatment of stereo-chemical information without the need of prior assignments.In effect, the optimization works on loose atoms and the topology is developed in the optimization process.Our method, called Conditional Dynamics, enhances the radius of convergence in refinement and may, in principle, be applied to random starting models for ab initio phasing.Calculations using a simplified test case, consisting of a polyalanine four helical bundle, showed i. a large radius of convergence even when the resolution is limited to 3.5 Å, and ii.successful ab initio optimization against 2.0 Å resolution data.Recently, we have developed a conditional dynamics-force field containing stereo-chemical restraints for all prevalent configurations in protein structures.We are now testing our method against real protein-diffraction data with respect to map improvement and ab initio phasing, i.e. modeling starting from random models.