On Fri, Oct 29, 2010 at 2:01 PM, Jianghai Zhu
If I use the riding H atoms come out of phenix.refine in molprobity, it gives me a decent clash score. But if I strip the H atoms and let molprobity to add H atoms, I get a much worse clash score. I saw the manual of phenix recommends the first method. So what is the cause of the difference here? I know long time ago that phenix and molprobity used different bond length to add the riding H atoms. I don't know what the situation is now. Does phenix.refine refine the riding H atom positions?
It won't refine them against X-ray data, but it will perform geometry regularization to idealize their positions - which also includes shortening the bond lengths to match what we observe by X-ray diffraction. The program Reduce (used by Molprobity and Phenix), on the other hand, uses the nuclear distances, which I think are on average 0.1A longer. The Richardson lab is working on standardizing and reconciling the different bond lengths, but it's a messy problem (especially when neutron diffraction is involved). -Nat