Re: [phenixbb] Low resolution refinement

GMail let me down - the screenshot is actually attached now. [image: Inline image 1] On Sat, Aug 23, 2014 at 9:44 PM, Nathaniel Echols wrote:

On Sat, Aug 23, 2014 at 3:00 PM, Appu kumar wrote:

...

I have a data from a crystal which we can process to 4.5A, and we are able to get a good molecular replacement model and it is a tetramer in ASU. I tried refining structure with phenix.refine but Rfree stuck at 0.41. I read a paper published by David Baker and phenix developer in 2013 about the low resolution refinement using a combination of rosseta and phenix. I would appreciate if i get a help on the refinement in rosseta along with phenix. This paper convince me that its worth trying the rosseta and phenix together for low resolution structure refinement.

I think it is worth a try, but I recommend (at the same time) that you look for potential explanations for the high R-free, and try other tools as well (starting with Xtriage, but also Tom's model morphing program) if you haven't already. The attached screenshot illustrates one example from the paper - it is a 4.0Å structure stuck at an R-free of 0.4, with very clear errors in the (sequence-idential) model reflected in the difference map. If your map looks like this, that may mean Rosetta et al. will help. Occasionally, it also means that the sequence identity is poor and you have to fix this before you and do more refinement.

For a more detailed answer, you should probably email us off-list at " [email protected]" with more details, log files, and, preferably, model and data. As always, these will be kept confidential within Phenix developers, but if you are not a P.I., please ask your P.I. if this is okay first! But I will also add some generic advice to the documentation - there is a fairly common set of diagnostics and procedures that get repeated at workshops or in other interactions with users, but I don't think are written down in any one place.

-Nat