Hi Jianghai,
Another thing about riding H atoms. As I said before, the bond distance between riding H and its attached atoms is different from phenix library (ccp4 mono library) to the library MOLPROBITY uses. I think the distance from MOLPROBITY is longer. The result is that the structure refined by phenix.refine has a not so good clash score. Is there a way to keep the distance from phenix.reduce in phenix.refine?
Yes, I remember you mentioned this in the past (may be even several times). And this is true: the distances in Mon.Lib. and internal library in Molprobity might be different. Theoretically, longer distances are more correct since they reflect the actual nuclear position (that one can observe in neutron diffraction experiment), while the X-rays show hydrogens shifted along the bond towards the atom they are bonded to. This is why some programs (like SHELX) use shortened distances for X-H bonds and some uses those obtained from neutron scattering. Overall, I agree that the dictionaries must be made consistent within the package - this is something to fix on PHENIX end. Pavel.