Dear Erron, We recently demonstrated that the high clash count typically seen with the conventional PHENIX refinement is likely attributed to a particular non bonded restraint function used by phenix.refine. When we employed QM/MM (ONIOM) refinement (Borbulevych et al., Acta Cryst., 2018, D74, 1063 https://doi.org/10.1107/S2059798318012913 ) that resulted in significantly better clashscore and improvements of other metrics. QM/MM refinement is available within PHENIX with the DivCon plugin. This plugin is available at no cost to academic users at http://www.quantumbioinc.com/products/software_licensing. Sincerely, Oleg Borbulevych Oleg Borbulevych, Ph.D. Staff Scientist QuantumBio, Inc. 2790 W. College Ave. State College, PA 16801 E-mail: [email protected], [email protected] Linkedin Professional Profile: https://www.linkedin.com/in/oleg-borbulevych-b8045422 -----Original Message----- From: "Titus, Erron" Sent: Friday, May 17, 2019 1:58pm To: "[email protected]" Subject: [phenixbb] minimizing clash score _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb Unsubscribe: [email protected], In order to improve model geometry of my x-ray structure, I am using real-space refinement together with reciprocal space refinement in phenix.refine. I like this, because it works very well to fix a lot of geometry issues. But clashes are a problem. Clashes tend to keep coming back. I found a clunky way to selectively use reference restraints to prevent most of the clashes from coming back, but I am wondering if there is another way of addressing this. I don't want to up-weight geometry globally, because 1) there are major regions of the x-ray map that are quite good, and 2) I don't have problems with other geometry, just clashes. I couldn't find much documentation on the "clash_guard" keyword... Thanks. Erron