On Fri, Aug 17, 2012 at 8:35 AM, Jeff Headd
We don't use an explicit C-beta positional restraint in Phenix, so there isn't any parameter you could change that would change its treatment as an outlier.
Technically, it will be restrained by a combination of the bond angle and chirality restraints. While it is in fact possible to adjust the weights on these for specific atom selections, I don't recommend it, for the same reason Jeff gives.
Much like rotamers, having an occasional outlier given very high resolution data is not a problem, and you can just ignore it if you believe your data supports the deviated chemistry. Validation metrics only reflect general guidelines for what is acceptable, and often true outliers can be some of the most interesting parts of a structure.
Another suggestion from Markus Rudolph (whose message to the list was bounced by the server): "is there a slight chance you have an alternate conformation of, say, 20% for this Thr? I have had this happen to me before and after inclusion of alternate conformations (based on density, of course), the Cbeta deviation went away." which is a very good idea, and quite common at high resolution (but very easy to miss). -Nat