Parameter sets

[Gerard 'CD' Kleywegt]

I've done a few quick slow cools to check if the Engh & Huber
force field is clearly better or not than parallhdg for
X-ray protein refinement.  Results follow:

--Gerard

--------------------------------------------------------------

* use crabp2 1.8 A model without ligand and waters

* run generate and strip all non-polar hydrogens for both

* check		WA xray		WA nmr
  8 - 1.8	149000		149000
  8 - 2.2	150000		148000
  8 - 2.8	151000		146000
  8 - 3.5	146000		138000

 -> use WA = 75000 for all jobs

* resolution	nr test refl	rel error in Rfree
  1.8		1475		2.6 %
  2.2		 826		3.5 %
  2.8		 381		5.1 %
  3.5		 177		7.5 %

* submit following job for both:
  - 8.0 - 1.8 A data
  - WA 75000
  - no electrostatics
  - weak harmonic restraints on CA atoms (force cst 20)
  - 100 steps RT MD without xray term
  - 50 steps Powell   ,,     ,,   ,,
  - 2000 K slow cool with xray term
  - 100 steps Powell  ,,

* R and Rfree:

		XRAY		NMR
  start		0.263 / 0.263	0.263 / 0.263
  after RTMD/Pw	0.358 / 0.359	0.344 / 0.338
  after SA/Pw	0.257 / 0.291	0.260 / 0.294
  error Rfree	~0.008

  -> differences within the estimated error !!

* repeat, using 8.0 - 2.2 A data:

		XRAY		NMR
  start		0.261 / 0.264	0.261 / 0.264
  after RTMD/Pw	0.348 / 0.346	0.334 / 0.328
  after SA/Pw	0.250 / 0.302	0.250 / 0.306
  error Rfree	~0.011

  -> differences within the estimated error !!

* repeat, using 8.0 - 2.8 A data:

		XRAY		NMR
  start		0.259 / 0.273	0.259 / 0.273
  after RTMD/Pw	0.332 / 0.341	0.316 / 0.316
  after SA/Pw	0.231 / 0.330	0.225 / 0.340
  error Rfree	~0.017

  -> differences within the estimated error !!

* repeat, using 8.0 - 3.5 A data:

		XRAY		NMR
  start		0.254 / 0.282	0.254 / 0.282
  after RTMD/Pw	0.305 / 0.314	0.297 / 0.308
  after SA/Pw	0.187 / 0.340	0.176 / 0.369
  error Rfree	~0.026

* ==> although Rfree is always lower for Engh & Huber than for
      allhDG, the differences are hardly significant (this is
      somewhat of a shock), except at 3.5 A.