[phenixbb] Twin_law in GUI Phenix Refinement

[Nathaniel Echols]

On Wed, Jan 25, 2012 at 1:29 AM, Tri Ngo <ngoductri at gmail.com> wrote:
> 1. I'd like to run phenix refinement with twin data but I don't know how to
> define twin_law using GUI Phenix refinement module in Linux.

Open the Settings menu, look for "Twinning", and there should be a
"Twin law" parameter in that dialog window.  In the nightly builds I
think I moved this to the main window, but I'll double-check.

> 2. I used phenix refinement in Windows to refine a 2.3A structure. My
> problem is the Rfree is quite high.
>
> - I applied twin refinement + TLS + NCS and SA refinement (wxu_scale = 1)
> Start: r_work = 0.2149 r_free = 0.2833 bonds = 0.016 angles = 1.988
> Final: r_work = 0.2029 r_free = 0.2735 bonds = 0.008 angles = 1.448
>
> - Then I use auto-weight to optimize the Rfree.
>
> optimize_xyz_weight = True
> optimize_adp_weight = True
> wxc_scale = 0.5
> wxu_scale = 1
>
> Start: r_work = 0.2152 r_free = 0.2735 bonds = 0.008 angles = 1.322
> Final: r_work = 0.2309 r_free = 0.2781 bonds = 0.003 angles = 0.848
>
> I think the bond rmsd and angle rmsd were too small in this final
> refinement. Do you think it's normal or I need to try more? Thank you very
> much for your helpful input.

I've found that weight optimization does tend to lead to lower RMSDs
than I'd been taught to expect; I'm not sure why this is.  They aren't
necessarily wrong - but this may be indicative of some other issue.
Do you have NCS?  TLS would also be worth trying, since it sometimes
helps with overfitting.

-Nat