[sharp-discuss] automatic MAD wavelength analysis confusion

[Clemens Vonrhein]

I'm not sure if 'inf' is enough ... try 'infl' (I should check).

Also: the f'/f" values should still follow the

  f"(peak) > f"(hrem) ~ f"(infl)

  f'(infl> < f'(peak) < f'(hrem)

order. autoSHARP tries hard to keep everything in synch - and
complains (as you noticed) easily. After all, with the f'/f" values
you have used before the SHARP refinement/phasing will start from very
wrong initial values - and who knows what could happen with those
precious few sites found in some below average MAD data.

Obviously, if you had really good MAD data all this wouldn't matter
too much - but autoSHARP is also intended to try and do as good as
possible on borderline cases.

See also:

  http://www.globalphasing.com/sharp/manual/autoSHARP1.html#needed
  http://www.globalphasing.com/sharp/manual/autoSHARP2.html#wave
  
Cheers

Clemens

On Tue, Aug 21, 2007 at 11:21:58AM -0400, Roni Gordon wrote:
> I did name the wavelengths "peak" and "inf", but then autoSHARP
> "changed" them even though I specified name, wavelength & f'/f''.
> 
> I thought that would be sufficient to disable auto-correction...
> 
> Roni
> 
> On Tue, 2007-08-21 at 16:27 +0100, Clemens Vonrhein wrote:
> > Hi Roni,
> > 
> > the 'problem' is that within autoSHARP we have two systems in place:
> > 
> >   - SHARP needs to know the f'/f" values of the heavy atoms
> >     (obviously) to do the refinement and phasing
> > 
> >   - SHELXC/SHELXD wants to know the 'names' of each dataset (PEAK,
> >     INFL, HREM, LREM) (presumably to translate these into some magic
> >     relation between the underlying f'/f" values ...)
> > 
> > So autoSHARP needs to keep track of both. In a good MAD experiment we
> > will have the situation that
> > 
> >   - peak has the largest f" value
> > 
> >   - infl has the lowest f' value
> > 
> >  - lrem/hrem are collected so far away from the edge, that the
> >    calculated values (via CCP4's CROSSEC program) are more or less
> >    correct
> > 
> > What happens often is that e.g. the lrem/hrem are collected 'just
> > remote enough' (but still fairly close to the calculated edge for
> > agiven element): this would confuse autoSHARP.
> > 
> > Or in your case it seems that you gave
> > 
> >   peak : f' = -8.0266  f" = 3.8421
> >   inf  : f' = -9.8034  f" = 3.8438
> > 
> > So here the 'inf' dataset has the lowest f' _and_ the highest f"! If
> > you want to give typical values I would do:
> > 
> >   peak : f' = -6.0  f" = 6.0
> >   infl : f' = -8.0  f" = 3.0
> >   hrem : f' = -1.5  f" = 3.5
> >   lrem : f' = -2.0  f" = 0.5
> > 
> > (or something similar). 
> > 
> > However, there is a little 'trick' (or maybe you could call it a
> > work-around) on top of these 'typical' values: give your wavelengths
> > some meaningful identifer ('peak', 'infl', 'hrem', 'lrem'). Then
> > autoSHARP should use those identifiers at the SHELXC/D stage - to keep
> > these programs happy.
> > 
> > Obviously, the fluorescence scan or jotted down numbers for f'/f"
> > would be best ;-)
> > 
> > Cheers
> > 
> > Clemens
> > 
> > On Tue, Aug 21, 2007 at 11:00:28AM -0400, Roni Gordon wrote:
> > > Hi all,
> > > 
> > > Just trying to re-process MAD data without the accompanying fluoroscence
> > > scans... I tried using "typical" values for peak & inflection (0.97924
> > > and 0.97940 A), and I was surprised when autoSHARP choked.  
> > > 
> > > Upon further inspection, both wavelengths were assigned a value of
> > > "peak" which was a problem for SHELX.  The following output was
> > > obtained:
> > > 
> > > 
> > >  WARNING : the identifier (peak) seems to suggest a peak wavelength, but the MAD
> > >            analysis results in assigning it as "inflection". We will be using the
> > >            user-defined assignement as "peak"!
> > > 
> > >              wavelength 1 (peak: 0.97924 Å with f'= -8.0266 f"=  3.8421 for Se) = peak
> > >              wavelength 2 ( inf: 0.97940 Å with f'= -9.8034 f"=  3.8438 for Se) = peak
> > > 
> > > The inflection is presumably still a few eV below the peak at this
> > > longer wavelength, and yet the tables still suggest a higher f''?
> > > 
> > > Any insight into this confusing (but I'm sure simple) matter would be
> > > greatly appreciated.
> > > 
> > > Roni
> > > 
> > > 
> > > _______________________________________________
> > > sharp-discuss mailing list
> > > sharp-discuss at globalphasing.com
> > > https://www.globalphasing.com/mailman/listinfo/sharp-discuss
> > > 
> > 
> 

-- 

***************************************************************
* 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)
***************************************************************