Re: [ccp4bb]: model bias in refmac5?

["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         ***

Jinsong Liu wrote:
> 
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
> 
> Dear All,
> 
> I would like to get some opinions from people using refmac5 about how
> serious the model bias the map coefficient (FWT, PHWT) would be.
> 
> I have a low resolution data, complete to 96% at 3.7A. Using molrep, I
> get different solutions from different homology models. Then I put these
> models in refmac5 for one cycle of rigid body (I got two molecules per
> ASU, each one as one rigid body) to calculate a 2fofc map. In most
> cases, the map looks too good to be true. But remember, these are all
> different MR solutions and only one can be real. I even throw in a
> totally different protein for Molrep, and refmac5, the map looks
> beautiful.
> 
> The R-factor for those refmac are around 51%-53% range, considered the
> model is only about 1/3 of the total size of the protein I am working on
> this R-factor is not too bad for this resolution.
> 
> Using exact same script, refmac (instead of refmac5) generated a worse
> map , though I can still see obvious model bias.
> 
> Using CNX model_map.inp, the calculated map is unthinkable for one case
> I tested, that is clearly not my solution based on the map.
> 
> In my memory, refmac map showed very little model bias except there are
> lots of missing data which is not the case I have here. This is the
> first time I saw such a heavy model bias in refmac. I used the standard
> script from the refmac5 example which is at the end of this mail.
> 
> Can anyone give me some hints?
> 
> Thanks
> 


 Garib should answer this but he is not here..

 You say your data only extends to 3.7A. At that resolution scaling (
and refinement) is tricky, and this can generate apparent bias. Look at
the plots of <Fobs> v <Fcalc> - they should more or less overlap.

 You can try turning off the bulk solvent scaling and/or the Babinet
scaling.

The best check is to set the occupancy for some conserved large side
chain to 0.00, and see if it re-appears in your map. If it does you can
be encouraged.

 Eleanor