On Mar 9, 2010, at 11:58 AM, Christian Roth wrote:
I did an anisotropic refinement with the gui for a 1.6 Ang. structure I think this is quite O.K. as the R-values seems to be reasonable(14/18.8). But for comparison I want exclude waters and hydrogens explicitly from the anistropic refinement. One could do this in the command line mode via adp.individual.anisotropic ="not water and not element H. I started from the gui and modified the .eff directly by icnluding: adp { individual { isotropic = None anisotropic = "not water and not element H" In the output pdb all waters have an anisou flag. So I assume phenix refined them anisotropically. How do I have to modify the .eff file to exclude the waters from the anisotropic refinement. I would like to use the gui, because you get this nice summary output with all statistics. Phenix started optionally anisotropic refinement at 1.7 Ang. As far as I understand one should keep in mind that the data to parameter ration should have certain value of app. 3 or 4. But restraints as paramters are not as solid as experimental data. Is this 1.7 Ang. borderline a rule of thumb, is there a strict rule at which point I could refine anisotropic, or maybe mixed isotropic anisotropic.
It isn't the R-values that matter, it's how much anisotropic refinement improves them. I always followed the advice in the SHELX FAQ: "Generally, if the drop in Rfree is less than 1 % you should revert to isotropic." This may be too strict, but the overall message is sound. You seem to have a rather large spread between R and R-free, which is slightly troubling (although not necessarily a deal-breaker). http://shelx.uni-ac.gwdg.de/~trs/shelxl_faq/shelxlfaq_index.html#Going%20Ani... As far as excluding waters goes, two suggestions: try explicitly setting the isotropic selection to the waters, and try using "resname HOH" instead of "water". (Although I thought "water" was supposed to work too...) Another approach would be to perform TLS refinement of the protein, which will incorporate anisotropy without greatly increasing the number of refined parameters. -Nat -------------------- Nathaniel Echols Lawrence Berkeley Lab 510-486-5136 [email protected]