> the liberty being offered is not so dangerous Parameter-wise, true, - if it is understood that you are now *modeling with little evidence*. How much electron density to affirm your model choices would one expect at say 7 ideal rotamers - which on top of static split will likely display dynamic motion?
On the other hand, the side chain HAS to be somewhere, and a distribution of the most probable rotamers with their respective frequencies may be more realistic than all other options (as long as the lack of specific experimental evidence is acknowledged). But overall, that model might be better due to the more reasonable prior expectation term although the R-value will be practically unaffected due to the lack of localized scattering contributions. Anyhow, if it can be abused, it will be ;-) Best BR -----Original Message----- From: Ethan Merritt [mailto:merr...@u.washington.edu] Sent: Tuesday, March 27, 2012 5:14 PM To: b...@hofkristallamt.org Cc: CCP4BB@jiscmail.ac.uk Subject: Re: [ccp4bb] REFMAC5 residues with bad geometry On Tuesday, March 27, 2012 04:35:40 pm Bernhard Rupp (Hofkristallrat a.D.) wrote: > >phenix.refine allows any number of alternate conformers. > > Hmm..... quoting our old friends from the validation circuit: Where > freedom is given, liberties will be taken.... True, but... [warning: back of the envelope calculation] Consider, for example, an isoleucine sidechain. It would require 12 positional parameters to refine the position of each sidechain atom (XYZ * {CB CG1 CG2 CD1}) for a single conformation. One the other hand there are only 7 rotamers in the library. So if you limited the refinement to the relative occupancy of the 7 ideal rotamers, that is a more parsimonious model than refining the individual atoms for one conformation. Season to taste with arguments about geometry restraints, but still I think the liberty being offered is not so dangerous. Ethan -- Ethan A Merritt Biomolecular Structure Center, K-428 Health Sciences Bldg University of Washington, Seattle 98195-7742
