If restraint weighting is 1/sigma^2 then setting sigma=0.0 may produce some interesting results. I suspected that setting them to zero might turn the restraints off but can't find a reference to support that. phenix.refine writes a .geo file that lists geometry restraints and sorts them within the file by residual (presumably weighted deviation). That's the first place I'd go look for outliers resulting from your definition of the geometry and compare residuals before/after refinement for what terms are being minimized at the expense of others. Phil Jeffrey Princeton On 2/28/17 11:44 AM, Maike Bublitz wrote:
Dear all,
thank you again for all your suggestions. Here’s what I’ve tried, without success:
1) reduce deviation from planarity restraints to very small value or zero -> ligand still gets distorted in refinement
2) Set all bond angle restraints to ideal values and checked sums of angles in planar rings -> ligand still gets distorted in refinement
3) Used phenix.elbow with - -final-geometry on a pdb of a planar ligand —> ligand still gets distorted in refinement
4) Used the grade server (grade.globalphasing.org http://grade.globalphasing.org) to generate cif file from SMILES string—> phenix.refine stops with message “Unknown file format: grade-ligand.cif”.
Does anyone have another suggestion, except for installing an extra QM package?
Best, Maike