CC1/2

From CCP4 wiki

Examples for citations from papers using (possibly among other criteria) CC1/2 and/or the paired refinement method (reference 1, below) for finding the high-resolution cutoff.

  1. "In choosing our highest-resolution cutoff we chose shells, which still fulfil CC1/2 (ref. 57) >25% and I/σ(I) greater than or equal to 0.5 and which were supported by our results from the paired refinement method (ref. 57). Choosing a lower resolution cutoff did not improve the overall quality of the maps." Models were refined at 6.0 and 3.4 Å, respectively http://www.nature.com/ncomms/2016/160419/ncomms11184/full/ncomms11184.html.
  2. http://www.nature.com/nature/journal/vaop/ncurrent/full/nature18934.html#methods "The final data set used for structure solution and refinement was merged from data from nine crystals for apo-SMOΔC and two crystals for vismo-SMOΔC. Data collection statistics are shown in Extended Data Table 1." That table shows high resolution values of 3.2 and 3.3 Å, respectively, with CC1/2 of 0.237 and 0.357, and I/σ(I) of 0.6. It also gives CC*, CCwork and CCfree and demonstrates that CCwork < CC*, as is to be expected.
  3. http://www.nature.com/nature/journal/vaop/ncurrent/full/nature19825.html This describes the results of an experimental phasing for XFEL data. The authors had the insight: "Handling of negative intensities was then implemented in cctbx.xfel, allowing the production of meliorated post-refined data sets with intensity distributions that no longer have abnormal reflection intensity metrics (L-test and NZ-test). We used the same criteria for the resolution cut-off, based on completeness on (>90%), redundancy (>4) and CC1/2 (>0.14); in all data sets, the resulting I/sigI in the highest resolution shells was greater than 0.5. While inclusion of negative intensities dramatically affects I/sigI, it results in stabilization of the refinement for the three native structures (pH 5, pH 7 and pH 10), requiring us to impose a lower weight on geometry in phenix.refine to obtain a structure with a plateauing Rfree. Inclusion of negative intensities in the data sets also resulted in a decrease of noise levels in q-weighted structure factor amplitude Fourier difference maps (see below). At the atomic level, the only noticeable change between structures refined with and without negative intensities included is an increase in the B-factors, reflecting the increase of the Wilson-B—and possibly giving a more realistic description of the BinAB structures at room temperature." Their findings seem to nicely support reference 2 (below).

References[edit]

  1. P.A. Karplus and K. Diederichs (2012) Linking Crystallographic Model and Data Quality. Science 336, 1030-1033
  2. Diederichs, K., Karplus PA. (2013) Better models by discarding data? Acta Cryst. D69, 1215-1222.