*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
Qun Wan, I think as a general rule of thumb after molecular replacement you should first evaluate your model, and make changes where necessary (for example, remove loops that poorly fit the density). Next I would say jump directly to the limit of your data in rigid body refinement. In this way your model avoids being overly refined before it is evaluated again by human eyes, and hopefully will avoid model biased refinement. As far as model validation goes, often errors are extended past the residues that are flagged, especially with ramachandran errors. Refinement programs sometimes can reasonable fix errors especially at high resolution, but at moderate to low resolution they tend to spread the errors because the model can fit the density well even with slightly distorted geometries. The best approach is to recognize and fix rotomer and ramachandran outliers early, before refinement has reach a minimum. Barring that, I'd say move your model over to CCP4 and run Refmac with tight geometry constraints. Then look around residues that were previously identified as outliers and see if whole sections can be tweaked to better fit the density. One program in particular that I'd like to advertise for re-fitting is King from the Richardson Lab. It has a tool called back-rub that rotates the Cb around the preceding and following peptide planes. The mis-alignment of Cb in my opinion often is the root cause of rotomer and even ramachandran outliers, because Cb sits on the cusp between backbone and sidechain. Careful evaluation of your Calpha geometry and the position of Cb can in my opinion surprisingly improve your model. With luck the tighter geometry constraints will help your R and Rfree converge, and hopefully drop. Cheers, -bob On 9/8/06, Qun Wan <[EMAIL PROTECTED]> wrote:
*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk *** Hi, I have a couple of questions during model refinement. Hopefully someone could help me. My crystal diffracted to 2.4Å resolution with molecular weight 40KD. The crystal belongs to P212121 space group with unit cell dimension a=72Å, b=73Å, c=88Å. The crystal is not twinned. After molecular replacement in the resolution range 20-4Å, I got a very good peak for later CNS refinement. In rigid body refinement, I am not sure which resolution range I should use, 20-4Å or 20-2.4Å? I tried both. If I use 20-4Å data, R=0.3355, Rfree=0.3265. If I use 20-2.4Å data, R=0.3304, Rfree=0.3207. After 20-4Å rigid body refinement, I also do phase extension gradually from 4Å to 2.4Å (that is, rigid body refinement 20-4.0Å, then 3.5Å, then 3.2Å, ...until 2.4Å), the result is R=0.3360, Rfree=0.3268. From the results of R and Rfree, I believe I should use 20-2.4Å in rigid body refinement. However, I was told to use 20-4.0Å first, then do phase extension like I said above. Could anyone tell me why? The second question is after final refinement and model building, my R=0.2133, Rfree=0.2333. It looks good. Then I validate the model by some other programs, like procheck, what If, and MolProbity. After a painstaking work, I am frustrated to find that my R=0.2056, Rfree=0.2398. I am confused why after rotamer correction, Ramachandran check, water check, collision check, etc., my Rfree increased, not decrease?! Your help is greatly appreciated!
