On Fri, Aug 10, 2007 at 12:32:21PM +0100, Eleanor Dodson wrote: > Sorry - Clemens is right. > > Eleanor
Hi Eleanor, phew - I was starting to wonder that what I put into (auto)BUSTER wasn't right. It should be trivial to put this into REFMAC too (Garib!): just add cards like FPRIme <atom-type> <f'-value> so that a user can do FPRIme Se -4.5 REFMAC then corrects C by the difference f'(CuKa)-(-4.5) (after reading the f'(CuKa) from atomsf.lib). And suddenly all those partially substituted Se-MET are becoming 100% substituted again ... a kind of 'in-silico expression system'. Cheers Clemens > > > Clemens Vonrhein wrote: > >Hi Eleanor, > > > >On Fri, Aug 10, 2007 at 10:27:16AM +0100, Eleanor Dodson wrote: > > > >>If you have significant anomalous scatterering and are using <F> you > >>need to modify the scattering factor for that atom appropriately. > >> > >>Here is an extract from $CLIBD/atomsf.lib > >>Se > >> 34 34 2.840900 > >> 17.000599 5.819600 3.973100 4.354300 > >> 2.409800 0.272600 15.237200 43.816299 > >> -0.879000 1.139000 -0.178000 2.223000 > >> > >>The values -0.879000 1.139000 are the f' and f" for Cu > >>wavelength ( 1.54A) > >> > >>I you think you have complete substitution and are using data at some > >>other wavelength for refinement you should change the f' and f" > >>appropriately > >> > >>( Copy $CLIBD/atomsf.lib to your area and edit it then assign refmac5 > >>ATOMSF my_atomsf.lib etc ) > >> > >>But in practice this is often less useful than you might expect and you > >>still see holes in your maps at the Se. I blame it on incomplete > >>substitution > >> > > > >Are you sure this would be the way to work it out in CCP4? As far as I > >can see (see top of atomsf.lib), the actual formfactor for an atom > >will be calculated as > > > > Formfactor: a1*exp(-b1*s*s) + a2*exp(-b2*s*s) + a3*exp(-b3*s*s) > > + a4*exp(-b4*s*s) + c > > > >with the coefficients (a1-4, b1-4 and c) given in atomsf.lib as > > > >AD This contains for each element: > >AD ID > >AD IWT IELEC C > >AD A(4) > >AD B(4) > >AD CU(2) MO(2) > > > >So I would be suprised if any program actually uses the tabulated > >f'/f" values for CuKa and Mo radiation to calculate formfactors. If > >you want to adjust the formfactors for elements that have an f' value > >very different from CuKa you need to change the C coefficient - > >according to the difference between the f'(CuKa) and > >f'(observed). This will have an observable effect then. > > > >At least that's what we do in BUSTER with good results ... > > > >Or am I missing something here? > > > >Cheers > > > >Clemens > > > > > -- *************************************************************** * Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com * * Global Phasing Ltd. * Sheraton House, Castle Park * Cambridge CB3 0AX, UK *-------------------------------------------------------------- * BUSTER Development Group (http://www.globalphasing.com) ***************************************************************