On Tuesday 09 December 2008 11:22:04 Joseph Ho wrote: > Hello, > > I have a refined 1.8A structure that I wonder if I could squeeze out some > anisotropy information. I did TLS refinement on the protein, and it helped my > Rfree. But I would like to ask a biological question based on the thermal > movement of only a few waters (7 total). In theory, that is not adding a lot > of parameters for anisotropy refinement because I am not refining anisotropy > of the whole protein. Correct me if I am wrong. So, > > 1) How do I use Refmac to refine anisotropic B's on only 7 waters? > There is an option to use "mixed" for B factor refinement, but I don't know > how to define the atoms.
You can just add ANISOU records for a selected set of atoms. Start with an isotropic description (off-diagonal terms = 0; diagonal terms equal to each other). But... > 2) If this can be done, given my resolution, how much confidence can I put in > my anisotropy > refinement of those waters? In another word, if I do see thermal ellipsoids > of the waters, > are they believable? And if I don't, is it because my resolution isn't high > enough to see it anyways? I would definitely not believe anisotropic refinement of individual waters at 1.8A resolution. Some while ago I ran tests by starting with a <1.0 A fully-aniso structure and gradually truncating the resolution of the data used for refinement. That way I could compare the resulting Uij values back against those obtained at high resolution. With sufficiently strong restraints, I could make the protein atoms more or less behave down to about 1.6A resolution, although the result was less good than using TLS refinement instead. However, by that point the Uij values for water atoms were not even mathematically stable, let alone reliably correlating with the "true" values. > I have a total of ~1700 atoms, and ~24000 unique reflections at 1.8A > resolution. > > Thanks! > Joseph -- Ethan A Merritt Biomolecular Structure Center University of Washington, Seattle 98195-7742
