[phenixbb] How to model domain-wide disorder in Phenix.refine (turn off bump interactions) ?

[Pavel Afonine]

Hi Phil,

the only way to do it phenix.refine is to assign different altlocs to 
overlapping atoms, say A to one domain and B to another one. Then they 
will not "see" each other via non-bonded interactions.

Also, in this case phenix.refine will refine occupancies of these atoms 
(which you can turn off if you wish). I guess by default it will group 
occupancies such that A+B=1 per residue frame (may be not, I forgot how 
I implemented it). That is for instance adjacent residues within A 
conformer may have different occupancies; same for B.
Perhaps in your case you don't want this: having all occupancies to be 
equal within each conformer would make more sense. You can do it by 
using this syntax:

refinement.refine {
     occupancies {
       constrained_group {
         selection = chain A and resseq 1:100 and altloc A
         selection = chain B and resseq 1:100 and altloc B
       }
     }
}

In this case all atoms in (chain A and resseq 1:100 and altloc A) will 
have the same occupancy between 0 and 1, and same for (chain B and 
resseq 1:100 and altloc B); and constraint occupacy_of(A) + 
occupancy_of(B)=1 will be used.

Let me know if you need any help with this.

Pavel

On 2/1/13 12:58 PM, Phil Jeffrey wrote:
> Hi,
>
> I've got a pernicious problem where the electron density for one 
> domain appears to extensively interpenetrate another similar domain.  
> I'm in the process of trying to work out things like: MR solution in 
> the wrong space group courtesy of pseudo-centering. In the meanwhile I 
> want to model it as interpenetrating domains, and need to turn off 
> VdeW bumps at the domain (chain) level of structure.  I can't just 
> model it as a infinite list of alternate locations.
>
> I (think) I remember how to do this in CNS (constraints interaction), 
> but is there a way to do this in phenix ?  I could not track down a 
> particularly promising section of either the manual or the .def/eff 
> file(s).
>
> Thanks
> Phil Jeffrey
> Princeton
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb