I am turning again to the wisdom of the community. In a paper, the authors used a dataset in P 1 2(1) 1 displaying tNCS (not mentioned anywhere in the paper) to model a dimeric structure. At the refinement stage, they used a twin law. Looking at the PDB-REDO entry, there is a warning line - "Used extra refinement runs to compensate for possible R-free bias". When I ran the data from PDB in Xtriage, I get the following:
1. translational NCS is present at a level that may complicate refinement 2. severe anisotropy (is this unsurprising because when I re-refined it, the CCwork/free didn't display a classical drop at the high resolution? artificial truncation?) 3. one or more twin operators show a significant twin fraction but since the intensity statistics does not indicate twinning, you may have an NCS rotation axis parallel to crystallographic axis 4. one or more symmetry operators suggest that the data has a higher symmetry 5. the intensity statistics look normal, indicating the data are not twinned. How should I proceed in re-refining this structure? Reindex to P222 and re-do molecular replacement? The reason I am interested in this model is that the authors claim their structure is a homodimer but PISA deltaG(diss) is around 4 kcal/mol. In a related structure where the protein was co-crystallized with another protein, the homodimer deltaG(diss) is about 7. Is this a significant difference? Thanks. Mohamed
