[phenixbb] problem with alternate conformations

Nathaniel Echols nechols at lbl.gov
Thu Jul 11 05:33:11 PDT 2013


On Wed, Jul 10, 2013 at 11:15 AM, George Phillips <georgep at rice.edu> wrote:
> I am trying to adapt my old CNS scripts to to multiple conformer refinements
> in Phenix, i.e. up to 20 copies of each entire chain.
> I seem to be OK with up to three copies of the whole chain, but when I try
> four or more I seem to be in an infinite loop in the 'reduce' phase of a
> refinement.
> 3 copies runs in less than hour, even with torsion simulated annealing, with
> four even overnight the program just runs 'reduce for hours and hours'.

This is not surprising, as alternate conformations aren't necessarily
considered truly independent copies the way an ensemble of models is.
(The reason being that usually the model will be a mix of single and
multiple conformations, especially where waters are concerned.)
You're handing Reduce a problem that it was never designed to deal
with.

> This
> happens on the standard release and development version 1422.
> Is there some way to turn off hydrogens as a work around?  Test files
> available on request. : )

Unfortunately the problem is that phenix.refine always (for the last
couple of years) calculates the clashscore as part of coordinate
refinement, instead of simply using the nonbonded restraints as a
target, and the clashscore by definition takes hydrogens into account.
 So there is no way to entirely avoid adding hydrogens if you are
using the individual_sites strategy.

-Nat


More information about the phenixbb mailing list