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

[ccp4bb]: Refinement vs twinned data



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

I am refining a model constructed  from MAD phases,  vs a native set of
twinned data treated with DETWIN. The data are merohedrally twinned in P61
& the twin fraction is estimated to be 0.3 using the Yeates server.

The native data go to 1.2A, but refinement doesn't seem possible beyond
1.5A ie. the residuals don't drop any further. The best model (with good
geometry, at least in REFMAC5) to 1.2 has an Rfree of ~21% and an Rwork of
18%. Is this a consequence of loosing some of the signal in the data due to
the detwinning process ?. I should mention I'm using TLS & CGMAT in
REFMAC5, refining isotropic Bs to 1.5 & thereafter refining
anisotropically. ARP/wARP was used to add waters, cycled with REFMAC5 once
the initial refinement stablised. I've tried dropping the resolution
steadily from 1.5 to 1.2 but this doesn't help either.

Also beyond 1.5A, REFMAC5 complains about waters having small eigen-values
when the model is refined anisotropically. Anitropic refinement at 1.5A
proceeds fine. I've tried using the DAMP 0.5 0.5 flag, but it doesn't help
& I'm not sure how I should alter the Sphericity restraint, which I though
might help - is this restraint dependant on the size of the structure or
the resolution of the data ?.

I'd really appreciate any comments/suggestions, esp. from those that have
experienced any of the above.

As always, thanks in advance.

Iain
*****************************
                       Iain Kerr
     Biomolecular Sciences Building,
                      3rd floor,                            
                 North Haugh,                        
                  St.Andrews, 
              The University,
              Fife KY16 9ST,                      
                Scotland, UK.       

        phone: +44 (0)1334 467265 

*****************************