Hi Paul, currently phenix.refine can refine occupancies as: individual, constrained individual and group. - In individual occupancy refinement occupancy of each atom is refined individually. - In constrained individual occupancy refinement occupancies of each atom are refined "individually" but it is made sure that the sum of them is one (for each atom, not group of atoms). - In group occupancy refinement you can refine one occupancy per selected group of atoms, without any constrains. What you mention is "constrained group occupancy refinement" where you refine one "individual" occupancy per group of atoms making sure that the sum of them is one. Unfortunately this is not implemented. This is on the top of my to-do list, so hopefully we will have it soon. Pavel. Paul Adams wrote:
Hi Pavel,
the occupancy should be the same for all of the atoms in an alternate conformation (grouped occupancy refinement). I doesn't make chemical sense for bonded atoms to have different occupancies.
Cheers, Paul
On Feb 6, 2008, at 8:56 AM, Pavel Afonine wrote:
Hi Christopher,
thanks for your questions!
We are working on a refinement at 1.14A which was suffering from lots of NPD atoms when refined anisotropically with refmac5 and/or shelx.
The implementation of anisotropic B-factors refinement in phenix.refine should never lead to non-positive definite ADP matrices or any "instability" in refinement.
1. In our model we have a few solvent ligands, and these show up in the log file as: Unusual residues: {'EDO': 11, ' CL': 1, 'SO4': 1} Classifications: {'undetermined': 13, 'water': 437} We've used a cif file for EDO with H's when needed, but still see this message. Is this normal? What is unusual?
Yes, it is normal. In this context "unusual" means that this is not "usual amino acid, water, dna/rna fragment".
2. When choosing "indivdual_occupancies" for the refinement, all the atoms within an alt-conf are refined to different occupancies- is this expected? It seems that all atoms in the A conf should have the same value, and the occ of all atoms in the B conf should also be the same. Instead, we see this: ATOM 232 CA AGLU A 28 26.163 7.163 4.422 0.77 6.73 C ATOM 233 CB AGLU A 28 24.851 7.269 3.672 0.80 9.27 C ATOM 234 CG AGLU A 28 24.978 6.767 2.259 0.66 12.91 C ATOM 235 CD AGLU A 28 23.662 6.717 1.554 1.00 16.51 C ATOM 236 OE1AGLU A 28 22.874 7.667 1.757 0.71 18.78 O ATOM 237 OE2AGLU A 28 23.429 5.733 0.803 0.60 16.77 O ATOM 240 CA BGLU A 28 26.140 7.156 4.424 0.23 6.33 C ATOM 241 CB BGLU A 28 24.794 7.230 3.713 0.20 6.48 C ATOM 242 CG BGLU A 28 24.877 6.766 2.280 0.34 6.55 C ATOM 243 CD BGLU A 28 23.536 6.702 1.598 0.00 7.53 C ATOM 244 OE1BGLU A 28 23.497 7.014 0.381 0.29 7.85 O ATOM 245 OE2BGLU A 28 22.543 6.341 2.280 0.40 8.40 O
This is correct behavior and this is exactly what you should expect refining occupancies of atoms in alternative conformations: the sum of occupancies of conformation A and B must be one. So, in your example above:
ATOM 232 CA AGLU A 28 26.163 7.163 4.422 0.77 6.73 C ... ATOM 240 CA BGLU A 28 26.140 7.156 4.424 0.23 6.33 C
the total occupancy is: 0.77+0.23=1 Same for other atoms.
3. Adding hydrogens during anisotropic refinement results in the Parvati server giving "Illegal Biso" errors for many of the hydrogens.
The H atoms at normal resolutions (not subatomic resolution) are treated as a special case. For example (default behavior), they ride on bonded atoms during coordinates refinement (riding model, distances X-H do not change), we refine one occupancy and isotropic B-factor per all H atoms in your molecule, etc. I don't know what exactly "Illegal Biso" means in Parvati server, but most likely you want to exclude H atoms for this analysis.
Just a suggestion: at resolution 1.4A you can try to change the default behavior for H refinement to this:
hydrogens { refine_adp = *one_b_per_residue one_b_per_molecule individual refine_occupancies = *one_q_per_residue one_q_per_molecule individual }
Please let us know if you have any questions or problems! Pavel.
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb