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

RE: high B-factors for some ncs copies

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

As a counter point to Phil's remarks, I've taken a structure from the
database and rerefined it with tight NCS and been able to achieve the same
Rvalue, better geometry, lower RMSD between models, and lower RMSD of B.
This was a case in which one monomer of a trimmer had a much higher average
B than the other two monomers.  'course this wasn't the case when I applied
strong NCS restraints on B.  I wasn't tracking FreeR at that time, so some
will discount this but I wonder how many people just aren't committed enough
to the idea of NCS.  People want to give up on NCS very quickly, or at least
loosen their restraints, saying that there is some real difference between
NCS copies of their molecule.  I think what their seeing is that it's more
difficult to get the Rvalue down to the desired value with strong NCS.

I have to remark on several specifics of Phil's priority list.

point 1)  The map can be misleading--I suppose everybody has heard of model
bias.  It is particularly important to be concerned 'bout this at 2.8
angstrom resolution.

point 2)  You should rely on features observed in your map to indicate that
the models deviate from strict NCS not on different outcomes from
refinement.  If you use strong NCS you should have reduced phase error and
better ability to see real differences in your maps.

point 3)  No complaints

point 4)  There is more than one example of a moderate resolution structure
with these kinds of freeR's and highly suspicious B's ( I was going to give
out a specific PDB code but I fear upsetting somebody).  It probably is true
that you have not mistraced your molecule with the kind of freeR you get,
are your B's well treated...

-----Original Message-----
From: owner-ccp4bb@dl.ac.uk [mailto:owner-ccp4bb@dl.ac.uk]On Behalf Of
Phil Jeffrey
Sent: Friday, June 25, 1999 1:17 PM
To: Susan Heffron
Cc: ccp4bb@dl.ac.uk
Subject: Re: high B-factors for some ncs copies

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

On Fri, 25 Jun 1999, Susan Heffron wrote:

> Chain name     Atoms       Bave       Bsdv       Bmin       Bmax
>      M1A        2894      33.57      18.77       2.00     100.00
>      M1B        2894      42.50      19.15       2.00     100.00
>      M2A        2894   ** 58.96 **   24.67       4.87     100.00
>      M2B        2894   ** 56.48 **   25.61       2.00     100.00
>      M3A        2894      34.52      21.01       2.00     100.00
>      M3B        2894      40.98      21.16       2.00     100.00
>      WAT          80      31.76      12.33       5.44      69.18
> Two copies of my molecule have alarmingly high average B-factors.

Not really.  I've had entire *structures* with that sort of average
B-factor.  Interesting, however that even when the <B> is 59 A^2 in one
chain there are some very low B-factors in the structure - is this
reasonable ?

I've seen this behaviour before - the "second" CDK2 molecule in
CDK2:CyclinA had a much higher <B> than the rest of the complex (2
complexes per asymm unit).  Doubtless there are many other examples.

> How should I check whether these high B-factors indicate a problem with
> chains m2a and m2b, or whether they simply have high B-factors?

In order of decreasing effort:

1.  Look at the density - does it look like it fits ?  If so, the <B> is
    probably a "real" effect.
2.  Do refinement without NCS restraints and compare the molecules - this
    may point out if the chain doesn't obey the NCS that well and you're
    just forcing it out of place (the B-factors just increase to flatten
    out the badly-placed atoms).  Compare with/without NCS models.
3.  If you are refining individual B-factors, is the increased B-factor a
    trend across the whole chain, or is it just in the loop regions (which
    are more likely to deviate from NCS) ?
4.  Your R-free is 23%, chances are you haven't made any large-scale
    errors like assigning the NCS inappropriately.

| Phil Jeffrey                                  |
| Crystallography Facility Manager              | If you lie to the
compiler, |
| Memorial Sloan-Kettering Cancer Center, NYC   | it will get its revenge
| phil@xray2.mskcc.org                          |     - Henry Spencer
| Voice: (212) 639 8547   Fax: (212) 717 3135   |