FreeR set reduced by half in twin crystal refinement?
Hi everyone, I have a P31 twin crystal, resolution 2.0, and with twin fraction 0.451. I use phenix.refine to refine the structure and with freeR set to 8%. % phenix.refine data.sca model.pdb twin_law="h,-h-k,-l" xray_data.r_free_flags.generate=True xray_data.r_free_flags.fraction=0.08 Result: Final: r_work = 0.1793 r_free = 0.2258 bonds = 0.011 angles = 1.557 REMARK 3 FREE R VALUE : 0.2258 REMARK 3 FREE R VALUE TEST SET SIZE (%) : 3.67 REMARK 3 FREE R VALUE TEST SET COUNT : 2013 After the refinement, it seems like my freeR set size reduced by half (3.67%). Is it due to the twin effect, or something is wrong with my refinement? And if I use ccp4 program "scalpeak2mtz " to convert the .sca to .mtz and with freeR fraction set to 8%. Then refine it with phenix. % phenix.refine data.mtz model.pdb twin_law="h,-h-k,-l" Final: r_work = 0.1669 r_free = 0.1970 bonds = 0.010 angles = 1.406 REMARK 3 FREE R VALUE : 0.2022 REMARK 3 FREE R VALUE TEST SET SIZE (%) : 7.80 REMARK 3 FREE R VALUE TEST SET COUNT : 4326 It turns out, my test set is still 8%, and R/Rfree looks even better than the first one. So I am wondering which one is the correct one and why? Thanks! Lauren
lauren wrote: (trimmed)
Hi everyone,
I have a P31 twin crystal, resolution 2.0, and with twin fraction 0.451. I use phenix.refine to refine the structure and with freeR set to 8%.
% phenix.refine data.sca model.pdb twin_law="h,-h-k,-l" xray_data.r_free_flags.generate=True xray_data.r_free_flags.fraction=0.08
Result:
Final: r_work = 0.1793 r_free = 0.2258 bonds = 0.011 angles = 1.557
After the refinement, it seems like my freeR set size reduced by half (3.67%). Is it due to the twin effect, or something is wrong with my refinement?
And if I use ccp4 program “scalpeak2mtz “ to convert the .sca to .mtz and with freeR fraction set to 8%. Then refine it with phenix.
% phenix.refine data.mtz model.pdb twin_law="h,-h-k,-l" Final: r_work = 0.1669 r_free = 0.1970 bonds = 0.010 angles = 1.406
It turns out, my test set is still 8%, and R/Rfree looks even better than the first one. So I am wondering which one is the correct one and why?
Thanks! Lauren
My first guess is that the R-free set generated in CCP4 is not set to follow twin laws, so you have work and free reflections associated by twinning. However, I would expect PHENIX to check this. As for the reduced size, that probably comes from the way the twinned test set is calculated. Maybe it is a bug. Are you using anomalous data? What is the PHENIX version? Joe Krahn
Hi,
a free set in phenix.refine is choosen such that
- you have at least n% of free flags, but not more then 2000+epsilon
(typically)
- it follows any possible twin law (not user specified, it follows symmetry
of the lattice)
the ccp4 suite doesn't follow the same rules, so you will
- have more free flags
- have twin related reflections in your mixed between the test and work set.
This reduces the rfree greatly and should be avoided.
P
2009/10/7 lauren
Hi everyone,
I have a P31 twin crystal, resolution 2.0, and with twin fraction 0.451. I use phenix.refine to refine the structure and with freeR set to 8%.
% phenix.refine data.sca model.pdb twin_law="h,-h-k,-l" xray_data.r_free_flags.generate=True xray_data.r_free_flags.fraction=0.08
Result:
Final: r_work = 0.1793 r_free = 0.2258 bonds = 0.011 angles = 1.557
REMARK 3 FREE R VALUE : 0.2258
REMARK 3 FREE R VALUE TEST SET SIZE (%) : 3.67
REMARK 3 FREE R VALUE TEST SET COUNT : 2013
After the refinement, it seems like my freeR set size reduced by half (3.67%). Is it due to the twin effect, or something is wrong with my refinement?
And if I use ccp4 program “scalpeak2mtz “ to convert the .sca to .mtz and with freeR fraction set to 8%. Then refine it with phenix.
% phenix.refine data.mtz model.pdb twin_law="h,-h-k,-l"
Final: r_work = 0.1669 r_free = 0.1970 bonds = 0.010 angles = 1.406
REMARK 3 FREE R VALUE : 0.2022
REMARK 3 FREE R VALUE TEST SET SIZE (%) : 7.80
REMARK 3 FREE R VALUE TEST SET COUNT : 4326
It turns out, my test set is still 8%, and R/Rfree looks even better than the first one. So I am wondering which one is the correct one and why?
Thanks!
Lauren
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
-- ----------------------------------------------------------------- P.H. Zwart Beamline Scientist Berkeley Center for Structural Biology Lawrence Berkeley National Laboratories 1 Cyclotron Road, Berkeley, CA-94703, USA Cell: 510 289 9246 BCSB: http://bcsb.als.lbl.gov PHENIX: http://www.phenix-online.org CCTBX: http://cctbx.sf.net -----------------------------------------------------------------
participants (3)
-
Joe Krahn -
lauren -
Peter Zwart