Hi Jianghai,

I tried "discard_psi_phi=False" in phenix.refine. There is only slightly improvement in the R, Rfree, and ramachandran plot (about 0.2% less disallowed residues, wxc_scale=0.2). A structure optimization (wxc_scale=0) with the phi-psi restraints has a bigger impact (about 2% less disallowed residues) with about 2% higher Rfree.

Thanks for your feedback!

What do you think is a good value for wxc_scale in a low-resolution refinement?

This is a tough one to answer. wxc_scale reflects how much you trust your data to guide you to the best model. That depends on many influences. I think the best - although expensive - approach is to simply try a range of values. Note that the esd on the psi-phi restraints are just someone's (?) best guesses. It may be worth playing with these values as well. For that, copy the data_link_TRANS, data_link_PTRANS, data_link_CIS, and data_link_PCIS blocks from mon_lib_list.cif to a new files, change the values, then specify the file on the phenix.refine command line. (You may want to double-check in the .geo file that your new values are actually used.)

What is the difference between wxc_scale=0 and phenix.geometry_minimization?

Should basically be the same, but in phenix.refine you can override parameters like discard_psi_phi. phenix.geometry_minimization doesn't allow you to do this (it is one of those many little things I never got to). Ralf