Re: [ccp4bb]: molecular replacement woes!

["Eleanor J. Dodson" <ccp4@ysbl.york.ac.uk>]

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

Yu Wai Chen wrote:
> 
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
> 
> > <FLAME>
> > So, basically I would like to strongly disagree.
> > </FLAME>
> 
> <NOFLAME>
> Pure scientific discussion.
> </NOFLAME>
> 
> > If you do have data at any resolution (the higher the better) and a
> > molecular replacement
> > model that you suspect it is correctly positioned, I would always use ALL
> > the data.
> > Thats your best chance to get the refinement going.
> 
> My comments came from experience, with current refinement programs, esp
> CNS, and not from old and ancient programs or computers.
> 
> First, density modification can blow up.  Second, B factor refinement can
> blow up.  These happened when I used high resolution data right away.
> Having said that, I always calculate maps using the highest-resolution data
> and use this to guide rebuilding.


My ha'penny

 Using all the data should improve your model more quickly providing
there is some modicum of truth in the solution.

 In my experience this is because the B values act as monitors - correct
bits keep sensible B values - the B values for loops in the wrong place
rapidly go to high values and their contribution is reduced. ( Make sure
you are not restraining the B s too stricyly though!) 
In other words B factyors SHOULD blow up for incorrect bits..

 Once that has happened the density modification (if you want to use it)
is more sensible too.

However if the model is wrong, or if there is bad overlap in the
solution with whole domains which need to shift a bit I agree the
refinement can go off the rails..

Then maybe rigid body refinement at lower resolution can help.


 All this relates to REFMAC5 rather than CNS but the ppls should be the
same
Eleanor