 |
 |
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Fri, 13 Feb 2009, Andy Millston wrote: > I have some very basic questions about model refinement process. > > 1. Why Refmac5 refinement from Coot GUI gives different results than > Refmac5 refinement from CCP4i? One obvious difference is that COOT > refinements are local whereas CCP4i refinements are global. But why > should it make a major difference? refmac5 is started with quite a few parameters, not just the input files. The default parameters from coot may differ from those from the ccp4i GUI. You should check the beginning of the corresponding log-files to find the differences. > 2. Is it acceptable to deposit a model to PDB which has been last > refined using Refmac5 from within COOT GUI? As long as your model is acceptable with respect to your data it doesn't matter how the model was derived. However, the word 'acceptable' hides a great amount of knowledge and experience. To start, you could e.g. read G. Kleywegt's latest article in ActaD: "On vital aid: the why, what and how of validation", Acta Cryst. D (2009) Vol 65, part 2, pp 134-139 > 3. Which cases warrant use of TLS and NCS restraints? When is it most > appropriate to use anisotropic refinement? As a coarse rule of thumb, you could use TLS between 1.6A and 3.5-4.5A resolution (beyond 4A resolution refinement becomes tricky anyhow, no matter what strategy/ programme you use). NCS is good at the beginning of refinement when the model is still not in good agreement with your data. or at resolution worse that, say, 3A - it depends on your data to parameter ratio. at better than 1.6A resolution I would start considering the use of anisotropic refinement. But beware: these are rules of thumb and may differ from case to case, depending on the data quality and crystallographer's experience. > 4. PHENIX permits defining different refinement strategies for different > parts of the structure. Can a structure part of which has been rigid > body refined in the last step, be deposited to PDB? Compare this question with the answer to question (2): without having tested this, I would think that if rigid body refinement makes ANY change to the position of your model AFTER it has been refined, it has NOT been refined properly. > 5. How to prevent Refmac from introducing close contacts during > refinement? Setting up geometric parameters (Distance restraint overall > wt 1.0) does not seem to be of any help. If you get clashes (close contacts) the first number I would look at is the WEIGHT between data and geometric restraints. It might be too high. Did you make refmac5 create hydrogen atoms? This might help reducing close contacts. Tim - -- Tim Gruene Institut fuer anorganische Chemie Tammannstr. 4 D-37077 Goettingen GPG Key ID = A46BEE1A -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) iD8DBQFJlZwFUxlJ7aRr7hoRAgHjAJ9cP2OkqfR6NXbkZZpSOdL3/jet3QCgh1dy OFrVVPE6EGVMJA1m5cozc2A= =JxpI -----END PGP SIGNATURE-----