Hi Andreas,
I guess you meant deconvolute the electron density maps. For me, it is a matter of taste. Twinned datasets are routinely detwinned (deconvoluted) since it is very difficult to build a protein in convoluted electron density. Likewise, the HIV protease people found they could more accurately fit the inhibitors using deconvoluted electron density.
If crystals with macroscopic twinning are detwinned, why not detwin crystals with microscopic twinning? It makes life for the non-crystallographer users of the pdb much and much easier. Mathematically it is equivalent and only a different way of presenting the data. One could always put in the publication a picture of the convoluted electron density. But this is only my humble opinion.
Herman
-----Original Message-----
From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Andreas Förster
Sent: Thursday, January 29, 2009 5:26 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] Small lines in diffraction pattern (more info)
Ian and Herman,
does one want to convolute the electron density at all? I was under the impression that current thinking favors convolution of the model instead, i.e. placing both the helices in both orientations at partial occupancy and letting the refinement program figure things out?
Andreas
[log in to unmask] wrote:
> Dear Ian and Margriet,
>
> You are right, the correction needs to be done on F, not on |F|^2. If I recall correctly (I did not do it myself), the assumption was that Fobs = 0.5*Fobs(A) + 0.5*Fobs(B), so Fcorrected(A) = 2*Fobs - Fcalc(B) where A and B are the two orientations. Since one does not have an observed phase, one would have to take calculated phases. I am unsure though, if that was done in practise and one did not just subtract the absolute values. Since the inhibitor is usually only a small part of the total scattering mass, the phases might not differ too much and therefore the error would not be too big. In case of superposition of base pairs, I guess that the differences in scattering between the different base-pairs is not too much, so one might also be able to get away with not using phases, but here you are the expert.
>
> Using this method, one could much better interpret the convoluted electron density, but one has to be very careful not introducing severe model bias. I would look in the literature in detail, what people from the HIV protease field had done to solve this problem.
>
> Cheers,
> Herman
>
--
Andreas Förster, Research Associate
Paul Freemont & Xiaodong Zhang Labs Department of Biochemistry, Imperial College London
|