Dear Phenix Guys, I have recently been trying my hand at the ensemble refinement option in phenix.refine. I generally observe an increase in rfree after performing the ensemble refinement with a well refined (using anisotropic B) structure, but the r-free increases are fairly small. For instance, 18.53% to 18.9% (1.8 Angstrom structure) or 20.45% to 21.1% (1.9 A)...I understand what this technically means (the ensemble refined model isn't as good) but given the noise in the calculation due to the limited number of reflections for the test set (~1000), would someone consider these rfree increases significant? My apologies if this question is a well-answered one...I can't seem to find much discussion about the range of rfrees to be expected or whether a 0.3-0.5% increase is large enough to toss out the refinement. Normally, I would take the lower rfree but because the ensemble refinement procedure gives potentially much more information, this seemed like a relevant question to ask. Thanks so much for your time, Arjun Raman Ranganathan Lab UT Southwestern ________________________________ UT Southwestern Medical Center The future of medicine, today.
Hi Arjun,
To assess the quality of different models it's important to look at a
range of metrics including but not restricted to Rfree. These include
Rfree:Rwork ratio, geometric indicators etc. I'm sure others can
advise you further here.
For the ensemble refinement specifically, just to check, did you try
arraying the pTLS and bath offset? This can have a big effect, for
more info see here:
http://www.phenix-online.org/version_docs/dev-1358/ensemble_refinement.htm
All the best,
Tom
On 19 December 2013 15:25, Arjun Raman
Dear Phenix Guys,
I have recently been trying my hand at the ensemble refinement option in phenix.refine. I generally observe an increase in rfree after performing the ensemble refinement with a well refined (using anisotropic B) structure, but the r-free increases are fairly small. For instance, 18.53% to 18.9% (1.8 Angstrom structure) or 20.45% to 21.1% (1.9 A)...I understand what this technically means (the ensemble refined model isn't as good) but given the noise in the calculation due to the limited number of reflections for the test set (~1000), would someone consider these rfree increases significant? My apologies if this question is a well-answered one...I can't seem to find much discussion about the range of rfrees to be expected or whether a 0.3-0.5% increase is large enough to toss out the refinement. Normally, I would take the lower rfree but because the ensemble refinement procedure gives potentially much more information, this seemed like a relevant question to ask.
Thanks so much for your time,
Arjun Raman Ranganathan Lab UT Southwestern
________________________________
UT Southwestern Medical Center The future of medicine, today.
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
-- Dr Tom Burnley, PhD Science and Technology Facilities Council (STFC) The Research Complex At Harwell Rutherford Appleton Laboratory, R92 OX11 0FA 01235 56 7871
Hi Arjun,
Our recent paper in Structure makes an estimate of the noise in refinement by asking what happens if you assign two different Rfree sets and then refine the same structure with 256 different refinement strategies. Based on that, I would say 0.5% difference in Rfree is within the noise range, and you might need 1% change to be sure it is meaningful. However, some structures showed more noise than others with this procedure. I would second the suggestion to consider other metrics of quality in addition to Rfree.
Kendall Nettles
On Dec 19, 2013, at 10:27 AM, "Arjun Raman"
participants (3)
-
Arjun Raman -
Kendall Nettles -
Tom Burnley