|Topic discussion: HKL2000|
The NIH X-Ray Diffraction Interest Group
web site: http://mcl1.ncifcrf.gov/nihxray
HKL2000 has a built-in feature (Reprocess) to reintegrate raw date (frames) with unit cell parameters fixed at the values obtained after postrefinement using Scalepack. Any comments, experience, or recommendations?
Xinhua Ji (NCI): Dr. Lothar Esser pointed out that the unit cell parameters might change during data collection due to dehydration, radiation damage, etc. I discussed another HKL2000 option "integrate with fixed distance" with Professor Wladek Minor at the recent Mid-Atlantic Macromolecular Crystallography Workshop and realized that the crystal-to-detector distance might change as well because the irregular shape of most macromolecular crystals could result in the imperfect centering of the crystals. Therefore, the real center of a data crystal could periodically move along with the rotation of the spindle axis; allowing the fluctuation of crystal-to-detector distance is in fact realistic. Wladek emphasized that the most important issue during integration with HKL2000 was the accurate prediction of the spots; it is recommended to fix parameters only when necessary for the improvement of prediction.
Lothar Esser (NCI): I just read Dr. Garboczi's Topic Discussion on HKL2000 vs XDS. As you said there may be something I can add in particular with regard to absorption correction. But let's go through the items in order.
My take on refining cell dimensions in Denzo is this that Denzo's job is to locate each spot and integrate the intensity that belongs to a particular Miller index. This is done by predicting the location of each spot based on crystal orientation, a variety of geometric factors of the experimental setup and of course on the current cell dimensions. If we were to say that cell dimensions are not to be refined (other than the necessarily fixed ones in a given crystal system) 'because a crystal can have only one fixed set of cell dimensions', we are shifting the burden of keeping a good match between observations and predictions to all other refinable parameters. Can this be done ? In an ideal world the answer is "yes", in reality the answer is "perhaps in some cases". So are cell dimensions changing or is the refinement of them only a means of inflating the number of parameters (over fitting?)
My personal view on this is, that cell dimensions can change for real.
1) Crystal decay. There is no reason to believe that a crystal that decays will doggedly cling to its initial cell dimensions. Effects like local dehydration, decarboxylation, bond cleavage and more come to mind that could alter the cell.
2) Anisotropy. The diffraction pattern of an anisotropic crystal viewed from one side and another side may give rise to different cell parameters.
Again what I want from Denzo is the best possible estimate of the intensity/background that belongs to a particular Miller index. If this requires refinement of cell dimensions I have no objections. The final question is can Denzo do it ? It is the task then of the crystallographer to analyze the behavior of any drifts in the cell and intervene with fixing some parameters (like distance, crossfire etc) or even "Special Integration" if necessary.
Besides problems with twins and satellites, I can think of only one reason why you might need to fix the cell dimensions, and that is when a crystal is unusually accurately aligned and the diffraction pattern provides not enough variation in one dimension.
That said, in my experience, denzo_3d does a very good job overall even when mostly default options are being used including the refinement of cell dimensions.
(Quite frankly I worry more about what the correct spot radius/ ellipse might be if my spots are smeared or elongated in different directions in different parts of the image.)
About "Unknown Keyword: absorption": Your info_cr (license) determines whether you are allowed to use absorption correction or not.
The default license won't include Absorption Correction (zo3). Request a new license with this feature included.
David Garboczi (NIAID): I used to do this with the script form of denzo. I've read and thought that refining the cell during integration was a little suspect. After all, if one can't trust that the cell dimensions are fixed enough to define your crystal, maybe that crystal isn't worth it. Mosflm finds the cell dimensions first and then integrates. I would definitely recommend reprocessing and would like to know about the other buttons in HKL2000 For instance, our absorption correction button clicks like it works, but the logfile says "Unknown keyword: absorption".
We are using XDS a lot as well and like its complete and current documentation. Its radiation damage correction algorithm has helped our anomalous signal, such that I doubt we would have solved several structures without it.
This site is maintained by Dr. Xinhua Ji (firstname.lastname@example.org) on the NCI-CCR-MCL server (http://mcl1.ncifcrf.gov).