Hi Peter, when using torsion angle parameterization you need to use simulated annealing based refinement. Gradient-driven minimization (without using second derivatives) will not perform well. To enable it in your refinement: phenix.refine model.pdb data.mtz simulated_annealing_torsion=true I would recommend running a SA refinement in Cartesian space too - just for comparison: phenix.refine model.pdb data.mtz simulated_annealing=true Please let me know if you have any questions or would like to discuss this subject some more. Pavel.
I try to refine a big complex at low resolution. I am a little confused about the syntax of using torsion angels gradient-driven refinement. Is it now the default for individual sites refinement instead of cartesian ? If not, should I add : sites.Torsion_angles=all for example ? Could you shed some light on this,