Dear all, I am working on a low resolution dataset (4.2 angstrom) of a multidomain complex. Using MR and density modification with NCS I obtained map in which I am extending/modifying my MR-models for the different domains. I tried to build the mainchain as a polyala model (extending on my MR-models) and then refined my new model in phenix.refine using torsion angle refinement. I have fourfold NCS and I think a this point it would be advantageous to use NCS-constraints instead of restraints. Is this possible in phenix? If I set the coordinate_sigma in de .def file to zero will this result in constrained NCS-refinement? ncs { restraint_group { reference=XXX selection=YYY coordinate_sigma=0.00 b_factor_weight=10 } Thank you very much for your advice, kind regards, Kenneth
Hi Kenneth,
If I set the coordinate_sigma in de .def file to zero will this result in constrained NCS-refinement?
No, sorry, we don't have "strict NCS" in phenix.refine. It is on the list of things to do in the next grant cycle. Some background: phenix.refine uses gradient-driven local minimization, which minimizes a sum of a penalty terms of the form weight * penalty. Examples of "penalty" are (f-obs - f-calc)**2 or (distance_ideal - distance_model)**2, or differences between NCS-related coordinates. weight = 1 / sigma**2, so if sigma is zero you have a divistion by zero. Therefore we use sigma=0 as a flag to turn off the corresponding penalty term. A constraint system is generally much more complicated than a restraint system. Ralf
participants (2)
-
Kenneth Verstraete -
Ralf W. Grosse-Kunstleve