On the refinement of Se, I would suggest Nat's option 1 (refine Se occupancies) is probably a good approach, depending on exactly what is happening. I would avoid refining f' and f" in this case unless you have some reason to believe that the occupancy really is 1 and the f' and f" are not right. It is true as Nat mentions that f' and occupancy are correlated (the effective real scattering is the occupancy times the number of electrons minus f'), and you could get the same calculated Fc either way. However it is probably best to model the situation in a way that is close to physical reality as possible, so that the parameters you get are the most reasonable description of what is going on. All the best, Tom T On Apr 4, 2013, at 12:39 PM, Nathaniel Echols wrote: On Thu, Apr 4, 2013 at 11:22 AM, Georg Mlynek mailto:[email protected]> wrote: 1a. Providing the XDS_ASCII.HKL to phenix tableI works perfect, however there are no values written for CC1/2. The same file in phenix.merging_statistics I get the CC1/2. Am I missing something? No, I am. Sorry, will fix this ASAP. 1b. The only value phenix tablel one takes from INTEGRATE.LP (i guess CORRECT.LP will work too) is the wavelength - true? I'm not sure which specific files from XDS will actually work, but in general, yes, the wavelength is the only thing I extract from the integration log file. (Unfortunately this is the one parameter on which we really depend on logfiles, since the wavelength is - inexplicably - not reliably propagated in reflection files.) 2. I see some negative density around the seleniums in my selenomethionine labeled protein. I guess this is due to radiation damage. Is the accepted standard to do occupancy refinement for just the sidechain, or just parts of the sidechain, the whole residue or just leave it like it is. This seems like a matter of personal preference, but I would definitely *not* refine the occupancy of the entire sidechain unless I was modeling it as a mix of MET and MSE to reflect partial incorporation. So there are several options: 1) refine the occupancies for SE atoms alone 2) leave them alone 3) refine alternate conformations I prefer (1), but it's difficult to know exactly what the difference map holes mean; they could simply be Fourier artifacts. A less conventional option is to refine f' and f'' for the SE atoms, but leave the occupancies at 1. I'm still a little confused about how f' and occupancy are correlated*, but in my experience the anomalous refinement soaks up most if not all of the difference in occupancy. The problem right now is that you can't deposit these parameters in the PDB, but hopefully that will change soon. -Nat (* A follow-up question for more knowledgable folks: could the large dispersive difference at the K edge be reflected as apparent partial occupancy, or am I misunderstanding the effect on scattering?) _______________________________________________ phenixbb mailing list [email protected]mailto:[email protected] http://phenix-online.org/mailman/listinfo/phenixbb

On Thu, Apr 4, 2013 at 12:10 PM, Francis E Reyes

wrote:

Does phenix.refine substitute the wavelength dependent f'/f'' when computing f for atoms in refinement?

No - I added a wavelength parameter a couple of months ago, but that is used for ion picking, it doesn't change the input model. (Although some of us think that might be a good idea.) phenix.maps and phenix.find_peaks_holes *do* use the wavelength parameter in this way, however. (maybe it was read from a properly formatted mtz) MTZ files always contain a wavelength - the problem is that we can't be sure whether this was set correctly or not. I suspect that MTZ files produced by CCP4 processing tools will have the correct wavelength, but the downstream programs may choose to ignore this. (Most of our tools default to writing a wavelength of 1.0; I've started to fix this but it's going to take a while.)

Or does it assume CuKa scattering factors?

No, in phenix.refine atoms never become anomalous unless you specify the anomalous groups. -Nat