Hi,

I rarely disagree with Ethan, but there are in fact ways of getting some idea 
of how much of the ordered structure from your crystal is missing in your 
model.  It's not based on the difference between Fo and Fc but rather more on 
the correlation between them and how that varies as a function of resolution.  
How well this works depends very much on why your model is incomplete — in many 
circumstances Ethan will be right in practice!

Some time ago, I proposed the use of something called a SigmaA plot 
(http://www-structmed.cimr.cam.ac.uk/Personal/randy/pubs/a25349r.pdf).  SigmaA 
is a parameter that measures what fraction of the calculated structure factor 
is correlated to the observed structure factor in a particular resolution 
shell.  The correlation depends on both the fraction of the crystal that has 
been modelled and on the quality of the model.  Since the effect of errors in 
the model is greater at higher resolution, SigmaA is a function of resolution.  
The way it varies with resolution tells us something about the size of the 
errors in the model, and the limiting value at low resolution tells us 
something about the completeness of the model.  If we make some assumptions 
(most importantly that the errors are primarily coordinate errors that are 
drawn from the same isotropic 3D Gaussian distribution of errors for all atoms, 
and that the missing part of the structure has the same overall B-factor as the 
modelled part of the structure), then we expect a plot of the logarithm of 
SigmaA against 1/d^2 to be a straight line, where the slope is proportional to 
the mean-square coordinate error and the intercept is proportional to the 
logarithm of the model completeness.

By the way, the original SigmaA plot also assumed that the measurement errors 
in the diffraction data were negligible.  With the development of our new LLGI 
(intensity-based likelihood) target, we now have a way of correcting the 
estimation of SigmaA for the effect of errors, so we could do better if there 
was a resurgence of the SigmaA plot.  

We don't model bulk solvent nearly as well as ordered atoms, so these plots are 
not very linear at low resolution.  But the plots do tend to be surprisingly 
linear at higher than about 5-6A resolution.  If you're doing molecular 
replacement on a complex and you place a model for part of the complex, the 
intercept of a SigmaA plot computed before refining your model will usually 
give you a reasonable idea of how much is still missing.  Refining the model 
could make things worse because of the effects of overfitting your data.

However, the case of a complex was assuming that the missing part is as well 
ordered as the modelled part.  Frequently we have trouble completing a model 
because a domain or an NCS copy has much higher overall B-factors.  What that 
means is that the fraction of the scattering we can model also varies with 
resolution instead of being a constant, because the better ordered part 
accounts for an increasing fraction of the scattering at high resolution.  This 
violates the assumption about the coordinate error being the only thing that 
makes SigmaA vary with resolution.

So whether this might be useful to you will depend on where you are with the 
structure determination and why your model is incomplete!

Best wishes,

Randy Read

> On 21 Feb 2017, at 21:53, Hunter Moseley <[email protected]> wrote:
> 
> Is there a straight-forward way to estimate the amount of missing electron 
> density that a particular protein structure is missing based on the 
> difference between Fo and Fc?
> 
> It appears that the normalization of the Fc due to the employing of a maximum 
> entropy method that keeps Fo and Fc comparable to the standard deviation of 
> Fo would make this difficult. 
> Or am I missing something?
> 
> Cheers,
> Hunter
> 
> -- 
> Hunter Moseley, Ph.D. -- Univ. of Kentucky
> Associate Professor, Dept. of Molec. & Cell. Biochemistry / Markey Cancer 
> Center
> / Resource Center for Stable Isotope Resolved Metabolomics
>  Not just a scientist, but a fencer as well.
>    My foil is sharp, but my mind sharper still.
> ---------------------------------------------------------------
> Email: [email protected] <mailto:[email protected]> (work)   
> [email protected] <mailto:[email protected]> (personal)   
> Phone: 859-218-2964 (office)   859-218-2965 (lab)   859-257-7715 (fax)
> Web: http://bioinformatics.cesb.uky.edu/ <http://bioinformatics.cesb.uky.edu/>
> Address: CC434 Roach Building, 800 Rose Street, Lexington, KY 40536-0093

------
Randy J. Read
Department of Haematology, University of Cambridge
Cambridge Institute for Medical Research      Tel: + 44 1223 336500
Wellcome Trust/MRC Building                   Fax: + 44 1223 336827
Hills Road                                    E-mail: [email protected]
Cambridge CB2 0XY, U.K.                       www-structmed.cimr.cam.ac.uk

Reply via email to