[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [ccp4bb]: Conclusion of bulk solvent corrected fo-fc map (B-f actor restraints)



***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


I'm not convinced that the TNT-style B-factor restraints are the way to go.
The problem is that they only represent the observed B-factor distribution
averaged over all side-chains of a given type, and may well not be
appropriate in specific cases.  On average you find that for all residue
types the B-factor increases as you go outwards from the CA, and
consequently the TNT-style scheme applies a positive offset as well as a
weight.  The problem is that frequently the reverse behaviour is observed,
particularly when the terminal atoms of the side-chain can form H-bonds,
i.e. the B-factors actually decrease as you go out from the CA.

See http://people.cryst.bbk.ac.uk/~tickle/iucr99/s31.html for an example of
a TYR side-chain from a high resolution TLS refinement where the terminal OH
is H-bonded to another residue and therefore has a low B-factor, whereas the
main-chain is very floppy! (note that in this picture the colour of the
ellipsoid is related to the value of Biso, whereas the atom type is
indicated by the bond colour).  The motion of the main chain pivoting about
the TYR OH is thought to be responsible for the biological activity and if
we had applied the TNT-style restraint at an earlier stage of isotropic
B-factor refinement before we had the high-res data, we might have missed
this important information.

My own view about isotropic B-factor restraints is that they should be
B-factor dependent, instead of B-factor independent as now.  What is clear
from unrestrained refinements of high resolution structures is that
differences between isotropic B-factors of bonded atoms tend to be strongly
positively correlated with the mean B-factor of the pair of atoms, i.e. the
difference tends to be small when the B-factors are small, and large when
they are large.  This can be understood on the basis that it's only the
along-bond component of the displacement tensor that is actually restrained
by the bonding forces (at least for sigma-bonded systems, pi-bonded systems
may well behave differently).  The components perpendicular to the bond are
determined by libration effects and are restrained only by VDW and H-bond
forces.  Interestingly the differences do not correlate strongly with the
distance between the restrained atoms (i.e. whether it's a bond or a bond
angle restraint).

Analysis of the B-factor differences in unrestrained high-res structures
suggests a scheme of the following form:

	Wrestraint = Wo/(B1^4 + B2^4)   (where Wo is an adjustable constant)

i.e.  sigma(B) proportional to B^2.

We've found that use of this scheme can give significant reductions in R and
Rfree (for the structure described above with 2.0 Ang data, R fell from
0.122 to 0.101, Rfree from 0.249 to 0.164 and the difference Fourier was
much much cleaner when we switched from the old very tight PROLSQ restraints
to the new scheme - admittedly this data was atypically good and the model
well-behaved, more typically R & Rfree falls by ~ 1%, still with some
significant clean-up of the difference Fourier, relative to a Refmac
refinement using the current defaults).

Note that the effect of employing this kind of scheme is that the average
B-factor will inevitably increase - the effect of using inappropriate tight
restraints is to drag the large B-factors down towards the small ones which
of course tend to dominate the calculated structure factors.  Unfortunately
many referees tend to have a view of what a "good" average B-factor should
be that is based on old structures in the PDB which were refined using too
tight restraints - so be prepared to argue your case!

Cheers,

-- Ian Tickle


-----Original Message-----
From: Dirk Kostrewa [mailto:dirk.kostrewa@psi.ch]
Sent: 14 March 2003 07:51
To: CCP4BB
Subject: Re: [ccp4bb]: Conclusion of bulk solvent corrected fo-fc map
(B-factor restraints)


***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

I use the following weights which result from a molecular dynamics
simulation 
by Yu & Karplus (1985), Acta Cryst. B41, 191-201 in the old PROLSQ days. 
They've checked how sensible the default B-factor restraints (1.0,1.5,1.5, 
2.0) were and came basically to the conclusion that they should be lowered
by  
a factor of ~2 for main chain and a factor of ~3 for side chain restraints. 
>From then on I'm using the following restraints with good experience:

Main chain bond	2.0 A**2
Main chain angle	3.0 A**2
Side chain bond	4.5 A**2
Side chain angle	6.0 A**2

A probably better approach are the side chain specific B-factor restraints 
used in the later versions of TNT based on the observed B-factor
distribution 
in high resolution protein cystal structures by Tronrud (1996), J. Appl. 
Cryst., 29, 100-104.

Best regards,

Dirk.



On Thursday 13 March 2003 23:04, Mark DePristo wrote:
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
> The default values (for refmac5.1.24) are:
>
>   THERMAL FACTORS
>            Weight= 1.00
>       Main chain bond (1-2 neighbour)         1.50A**2
>       Main chain angle (1-3 neighbour)        2.00A**2
>       Side chain bond                         3.00A**2
>       Side chain angle                        4.50A**2
>
> The looser parameter set is enabled by:
>
> 	BFAC  1  5  6  6  7.5
>
> which produces:
>
>   THERMAL FACTORS
>            Weight= 1.00
>       Main chain bond (1-2 neighbour)         5.00A**2
>       Main chain angle (1-3 neighbour)        6.00A**2
>       Side chain bond                         6.00A**2
>       Side chain angle                        7.50A**2
>
> Mark
>
> On Thursday, March 13, 2003, at 01:44 PM, wgscott@chemistry.ucsc.edu
>
> wrote:
> > Please let us know what weightings you are now using.
> >
> >>  One definite conclusion (thanks to Ian Tickle) is that the default
> >> B-factor restraints in refmac are definitely too tight (and incorrect
> >> in the HTML documentation).  Loosening them up  ...
>
> Mark DePristo
> Ph.D. Candidate
> Dept. of Biochemistry
> Cambridge University
> mdepristo@cryst.bioc.cam.ac.uk
> http://www-cryst.bioc.cam.ac.uk/~mdepristo/

-- 

****************************************
Dirk Kostrewa
Paul Scherrer Institut
Life Sciences, OSRA/007
CH-5232 Villigen PSI, Switzerland
E-mail: dirk.kostrewa@psi.ch
Phone: +41-56-310-4722
Fax: +41-56-310-4556
WWW: http://www.sb.psi.ch
****************************************



CONFIDENTIALITY NOTICE 
This email contains confidential information and may be otherwise protected by law. Its content should not be disclosed and it should not be given or copied to anyone other than the person(s) named or referenced above.  If you have received this email in error, please contact I.Tickle@astex-technology.com.