CK wrote:
I think perhaps I should clarify what I've been doing. I've been refining my structure using phenix.refine and had reached a fairly satisfactory point with suitably low R-factors (21/25) and very reasonable RMSDs for bonds and angles. At this point I ran my structure through the molprobity server to take a closer look at geometry and clashes.
OK, so that's the issue, right there. Molprobity is a modern program - using PDB v3 atom names. Coot uses the Refmac dictionary, i.e. version 2.3 names. Coot does not recognise the new atom names e.g. C1' (in Ar) or HB3 (in ARG) and so pushes them out with non-bonded contact restraints.
If you want to use the output of molprobity with coot/Refmac dictionary, then you will need to downgrade the molprobity output file.
Yes, this needs to be fixed - upgrading and validating the new dictionary is a substantial task.
For the record, Coot ignores the occupancy of the atoms when refining. Adding hydrogens to a model should do reasonable (if not good) things no matter the resolution of the data.
As going through the entire clash list by hand can be very tedious I wanted to see if adding hydrogens in the riding positions and refining again would help alleviate some of the clashes, which it did.
Yes, it should. Paul.
