[sharp-discuss] Re: Heavy Atom Negative Density
Clemens Vonrhein
vonrhein@globalphasing.com
Wed, 30 Jan 2002 16:08:49 +0000
Hi Paul,
> Attached is a copy of my input file. It's setup just like the example - and
> I use the in-house data set as the reference.
Aha!!! Now things are clearer: you actually have a MAD + native + two
derivatives. Your setup looks like this
C-1 protein + Fe-S cluster(s)
X-1 ? crystal
W-1 in-house data
B-1
W-2 wavelength 1
B-1
W-3 wavelength 2
B-1
W-4 wavelength 3
B-1
C-2 protein + Hg
X-1
W-1
B-1
C-3 protein + Os
X-1
W-1
B-1
which is ok if
1. you measured your in-house data on the same crystal as your MAD
data
2. your 2nd and 3rd compound do not have Fe-S clusters
(one note: you want to take the best of your MAD wavelengths as a
reference wavelength to reduce corrleated non-isomorphism).
However, if
1. all your crystals contain the Fe-S clusters
2. the in-house was collected on a different crystal
the way to set it up is the following:
C-1 protein + Fe-S cluster(s)
X-1 MAD crystal
W-1 wavelength 1
B-1
W-2 wavelength 2
B-1
W-3 wavelength 3
B-1
X-2 in-house crystal
W-1
B-1
C-2 protein + Fe-S cluster(s) + Hg
X-1
W-1
B-1
C-3 protein + Fe-S cluster(s) + Os
X-1
W-1
B-1
At least that is what I suspect your chemical composition in the
various compounds are. The important bit is, that you have to declare
the Fe atoms in ALL compounds: although they are common to all
compounds, you have to declare them for the MAD - and therefore for
all others too.
In the two derivatives you then have to also declare the Os/Hg sites.
> No - I'll give it a try. However, why then do I get nice positive density in
> the residual map when I use just one Iron atom? Wouldn't these be negative
> too?
Residual maps don't use phases - they'll look identical for both hands
(well apart from inverted coordinates of course). you also don't need
to re-refine things in the inverted hand - just phase calculation will
do.
> The residual map from the 4Fe phased data set still has alot of strong
> positive density round the 4Fe-4S cluster . I assumed this was from the 4
> sulfurs and noise.
If nothing is AT the sites (positive or negative) then your f'/f''
seem to be ok. Beware when fixing the HAT_OCC if your data is not on
roughly absolute scale and/or your f' are wrong. And 3A is definitely
not going to be on absolute scale ...
Unless these features are still > 6 sigma I would ignore them -
otherwise you might want to put S atoms in - or model things with
anisotropic B or ...
> I used the EXAFS values for inflection and peak - and table values for
> in-house and remote. They were all kept fixed.
If you fix HAT_OCC you might want to refine f' values. But I rather
suggest fixing the f'/f'' values (if your EXAFS is good) and refine
HAT_OCC - even if you 'know' they should be 1.0 (assuming your data is
on absolute scale - which it never really is).
Some additional remakrs to your SIN file:
1. NANO_CLOC is HUGE for your in-house data and VERY large for your
MAD data: your residual maps are probably rather messy - or you've
missed several minor sites? Also, use one of the MAD wavelengths as
reference (see above).
2. how do you know that your Os and Hg derivatives have HAT_OCC of
1.0? I would always refine occupancies of soaks.
3. no anomalous data for the Os and Hg?
I would actually first stick with the 3 wavelength MAD, try to get the
residual maps clean etc. Then add another CRYSTAL (the in-house data)
below the exisiting one. A last resort would be to add the Hg/Os
compounds. But maybe/probably the MAD and/or MAD+native will be
enough?
Hope that helps
Cheers
Clemens
--
***************************************************************
* Clemens Vonrhein, Ph.D. vonrhein@GlobalPhasing.com
*
* Global Phasing Ltd.
* Sheraton House, Castle Park Tel: +44-(0)1223-353033
* Cambridge CB3 0AX, UK Fax: +44-(0)1223-366889
*--------------------------------------------------------------
* BUSTER Development Group (http://www.globalphasing.com)
***************************************************************