[phenixbb] clashscore

[Edward A. Berry]

I just tried this: increasing the nonbonded_weight just to 3000 reduced clashscore from 6.5 to 1.9. Also there were about half-dozen waters flagged by pdb as less than 2.2A from another atom - with the increased weight the closest was 2.30. At the expense of increaseing R-free by 0.0025 - perhaps not significant. It would be best to fix any clashes that result from errors, but I guess some atoms are poorly determined and wander about- a small percentage may wander too close to another atom, and the increased weight opposes that tendency.

On 11/27/2017 03:08 PM, Pavel Afonine wrote:
> Hi Pan,
>
> first off I suggest to inspect top worst clashes and see if they can be addressed. If that does not help, then try running several refinement runs each one using "nonbonded_weight" set to say 1000 (the default value), 5000, 10000, 25000, 50000. See how that changes the clashscore and model-to-map fit.
>
> I would appreciate if you tell us what you find.
>
> If you are using the command line then it is
>
> phenix.real_space_refine model.pdb map.ccp4 resolution=3.1 nonbonded_weight=10000
>
> In the GUI:
> Refinement Settings -> All parameters -> Search parameters : type into prompt nonbonded_weight, then OK -> In pop up prompt change 1000 to a desired value.
>
> Pavel
>
>
> On 11/22/17 17:19, 潘 晓伟 wrote:
>> Hello everyone
>> I refine a cryo EM data with 3.1 angstrom, the model fit well with the map, but the clashscore is high, abot 23.
>> How can I decrease the value of clashscore with realspace.refinement in Phenix GUI.
>> Thanks!
>> Pan
>
>
>
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb
> Unsubscribe: phenixbb-leave at phenix-online.org
>