[phenixbb] Rfree refmac5 issue

Nathaniel Echols NEchols at lbl.gov
Wed Dec 16 01:46:31 PST 2009

On Dec 16, 2009, at 9:12 AM, anon wrote:
> has anyone had any experience using phenix with refmac5? I seem to be having issues with the Rfree flag. Phenix sets aside 8.1% of the reflection for Rfree. However when i use refmac for refinement, it reports that its using 91.4% of the reflections for Rfree........?

I assume you let Phenix pick your R-free flags for you.  CCP4 programs (almost?) always expect that the R-free flag is 0, even if that means that 90% of your reflections end up in the test set.  You need to tell Refmac to use 1 instead - if you're using keyword input, this means


If you download the latest nightly build, the reflection file editor GUI has an option "Convert R-free flags to CCP4 convention" (click the "R-free flags generation" button, and set user level to Advanced).  This is a massive hack, and definitely a use-at-your-own-risk feature, so I recommend inspecting the file it outputs to make sure it does what you expect.   (It will probably be sufficient to run phenix.model_vs_data - or the validation tools in the GUI - to confirm that the R-free is the same with both the input and output files, since model_vs_data will automatically detect the CCP4-style flags and switch to using 0.)  If you run into problems, feel free to email me separately.

Nathaniel Echols
Lawrence Berkeley Lab
NEchols at lbl.gov

More information about the phenixbb mailing list