[sharp-discuss] Heavy atom density

Clemens Vonrhein vonrhein at globalphasing.com
Wed May 13 07:55:29 CEST 2015

Hi Lothar,

a very interesting set of questions ...

On Mon, May 11, 2015 at 10:00:52AM -0400, esserlo at helix.nih.gov wrote:
> In my SIRAS case, with a good native but still better Hg derivative data set
> to resolutions of about 3.7 and 3.3 A respectively, I do get pretty good
> experimental phases with Auto Sharp 2.8.1 SUSHI 3.10.1. So I am
> happy.

Very good :-)

> However in density modification and later on in refinement that
> includes experimental phases (HL coefficients) I do get petty strong
> features for the Hg atom sites.

Those HL coefficients are related to the unperturbed structure factor
FP/SIGFP. In your case it will represent two slightly different

 * up to 3.7A (and for all reflections with both Fnat and Fder
   measurements) it will be the unperturbed structure-factor amplitude
   of the reference compound.

   You probably did run with the native as first compound and the
   derivative as second - in which case it is the Fnat that acts as a
   reference here.

 * beyond 3.7A (only Fder measurement) it will be the unperturbed
   structure-factor amplitude of the Hg derivative.

So it might be better to run SHARP with the Hg dataset as
first/reference compound and Fnat as a second compound. This way your
FP/SIGFP and HL will all relate to the unperturbed structure-factor
amplitude of the Hg derivative over the whole resolution range.

> Since there are bulky residues in the vicinity, there is a tendency for them
> to move into the strong ghost sites of Hg. Without experimental phase support,
> things clear up (no pseudo Hg density, obviously) but I would say that overall
> refinement clearly benefits from having experimental phases around.
> {Refinements were done in BUSTER and Phenix with pretty much the same effect}
> Maybe I am picking up the wrong coefficients ?

Are you refining against the FP/SIGFP from SHARP (in eden.mtz) or
against the F/SIGF for the native dataset?

> I did use HLA-D from Sharp which are the description of the
> calculated phase probability and my understanding is that this is
> the right set to use.


But remember: the definition of the radius of the "mean circle"
(resulting in FP/SIGFP) does not completely eliminate the cross-talk
between phase error and amplitude error.

> I am aware of the documentation that speaks of phases with average
> heavy atom contribution like PHIBshasol but there is also
> PHIcentshasol which is "light atoms only". I tried all of them (with
> and without average heavy atom contribution ) in ccp4 DM:

DM won't take the offset (by the FH/PHIH, ie Hg atoms) into
account. The SOLOMON-driving script in SHARP/autoSHARP does take that
into account though.

But what is seen in the resulting maps depends a bit what FP/SIGFP
represents (see above) and how the density-modification tool handles
density within the 'protein' region that is higher than the assumed
standard histogram of density.

For some discussions see:


> Density modification and averaging clearly have beneficial effects,
> but I still see pronounced density stemming from Hg. No matter what
> I do.
> Well, is this to be expected that even the "light atoms only" phase contain
> information about the position of Hg atoms?

The HLA-D are not light-atoms only, only the Fcent/PHIcent
coefficients (for map computation) are - if the reference compound in
SHARP is the native (see above).

> Any suggestions ?

Maybe you could refine the Hg-containing structure including the HLA-D
coefficients first (after running SHARP with the dataset in order
Fder-Fnat): this is the higher-res dataset anyway. Then move to the
lower-res native:

 * starting from the higher-res Hg structure (but without the Hg

 * using BUSTER with the -target option to do LSSR-style restraints of
   the model in the low-res datasets against the final/best model in
   the higher-res one


Clemens, Gerard & Claus (for sharp-develop)


* 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)

More information about the sharp-discuss mailing list