Thank you Nathaniel! But through the GUI, the program can run smoothly with twin refinement. My only problem with the GUI is that I cannot find the the place to generate R-free flags in the highest symmetry for twinning refinement. Thanks again, Chen On Mon, Apr 21, 2014 at 2:57 PM, Nathaniel Echols wrote:

On Mon, Apr 21, 2014 at 10:40 AM, Chen Zhao wrote:

...

Thank you all for your help and the program is running now. However, I met a AttributeError as " 'twin_model_manager' object has no attribute 'f_model_scaled_with_k1' ". Sorry for so many problems.

This sounds like the operation of the program in the default mode is incompatible with twinning.

-Nat

Dear Nathaniel, Thank you for your reply. I am trying to run it with simulated annealing refinement. Because I am a new user I might have made some mistakes, but the error message popped up during the second cycle of twinning refinement at the beginning. For the R-free, I agree that I could use the data generated by phenix.refine (because I generate the R-free in phenix.refine). Sincerely, Chen On Mon, Apr 21, 2014 at 3:45 PM, Nathaniel Echols wrote:

On Mon, Apr 21, 2014 at 12:39 PM, Chen Zhao wrote:

...

But through the GUI, the program can run smoothly with twin refinement.

It's not clear what you mean - the operation of the program should be no different on the command line versus the GUI. All that matters is what method you use for the omit map preparation. The default "simple" method ("iterative" might be a better name for it) is apparently incompatible with a twin law. The "refine" or "anneal" methods will work fine with a twin law, but are far slower even if you have a big cluster. What exactly are you trying to run?

My only problem with the GUI is that I cannot find the the place to

...

generate R-free flags in the highest symmetry for twinning refinement.

The lattice symmetry option is already enabled by default. However, if you already have R-free flags defined, it will not generate a new set; for this, you need to use the reflection file editor (GUI->Reflection tools). However, if you're at the stage of generating a composite omit map, this usually means that you've already done refinement and have R-free flags already - in which case calculating maps or performing further refinement with a different test set is a bad idea. If you want to change test sets in the middle of the refinement process, you should first do some simulated annealing refinement to un-bias R-free.

-Nat