Andre, some suggestions:
1. First be sure that shape reconstructions as an approach to
interpreting your data is justified – the shape reconstruction approach can be
quite misleading in some circumstances:
i. Does
your data show any evidence of flexibility and lack of compactness from Kratky
and Porod-Debye analyses? Such aspects can exaggerate molecular volumes,
undermining the ability to implement shape reconstruction algorithms to arrive
at stable solutions. Extra floppy bits like extended linkers his-tags and
mixtures of conformations can also affect apparent volumes.
ii. Does your
data agree with your understanding of molecular mass at those concentrations
(e.g., Mass by Qr, Mass by I(0), Mass by empirical relationships with Porod
Volume, etc.). Does Oligomer/Mixture analysis tell you that you have all
tetramer, rather than say 95% tetramer/5% dimer…? Is there evidence of
aggregation or concentration-dependence through multiple concentrations? As one
can guess, mixtures and aggregation can undermine reliable shape reconstruction
and interpretation.
iii. How does
the atomic inventory of your model compare to that of the reconstruction?
Commonly, x-ray crystal structures are missing sequences that might otherwise
be in a full-length/native construct. If your scattered sample has a
composition not entirely represented in your atomic model, automated approaches
might be misled, and doing it by eye might be difficult without obvious
landmarks or constraints. It could be very helpful to scatter a few different
truncations and then to employ simultaneous solution approach used in MONSA.
Increasing your data-to-parameters always helps!
iv. How does a
CRYSOL/FOXS fit between your model and the primary data look, independent of
the shape reconstruction calculations?
v. Does a
symmetry-free (P1) calculation using DAMMIF or GASBOR agree with a
symmetry-imposed calculation?...What is the distribution of the Normalized
Spatial Discrepancies (NSDs) and Chis like for 10+ calculations? Is averaging
justified by the statistics?
2. With regards to software suggestions, in addition to SUPCOMB, I might
suggest looking at the SITUS/SCULPTOR package. It uses a real-space approach
to reconcile atomic models with volumetric representations, and does provide a
real-space correlation coefficient for fits.
Hope that helps,
Kushol
Kushol Gupta, Ph.D.
Research Associate - Van Duyne Group
Department of Biochemistry and Biophysics
Perelman School of Medicine at The University of Pennsylvania
<mailto:[email protected]> [email protected] / 215.573.7260 / 267.259.0082 /
<http://www.stwing.upenn.edu/~kgupta> www.stwing.upenn.edu/~kgupta
From: CCP4 bulletin board [mailto:[email protected]] On Behalf Of Andre
Godoy
Sent: Monday, February 2, 2015 5:53 AM
To: [email protected]
Subject: [ccp4bb] [off topic] Fitting unknown model in SAXS envelope
Dear users
I'm having some troubles to fit my x-ray model in my SAXS envelope..
more about:
1) I have a SAXS model with enough room for 6 monomers.
2) I have the crystallographic structure, but AU or any generate symmetry
related doesn't appears to be the biological unit (I mean, crystal packing is
different from SAXS packing)
Is there any piece of software that can take monomers and find the best (or
least worst) RMSD between a SAXS envelope and a generated coordinate system? Or
anyone have a good ideia for me to do so?
All the best,
Andre Godoy
PhD Student
IFSC - University of Sao Paulo - Brazil
