DIBER: protein, DNA or both?

The program DIBER (an acronym for DNA and FIBER) requires only native diffraction data to predict whether a crystal contains protein, B-form DNA or both. In standalone mode, the classification is based on the cube root of the reciprocal unit-cell volume and the largest local average of diffraction intensities at 3.4 A resolution. In combined mode, the Phaser rotation-function score (for the 3.4 A shell and a canonical B-DNA search model) is also taken into account. In standalone (combined) mode, DIBER classifies 87.4 +/- 0.2% (90.2 +/- 0.3%) of protein, 69.1 +/- 0.3% (78.8 +/- 0.3%) of protein-DNA and 92.7 +/- 0.2% (90.0 +/- 0.2%) of DNA crystals correctly. Reliable predictions with a correct classification rate above 80% are possible for 36.8 +/- 1.0% (60.2 +/- 0.4%) of the protein, 43.6 +/- 0.5% (59.8 +/- 0.3%) of the protein-DNA and 83.3 +/- 0.3% (82.6 +/- 0.4%) of the DNA structures. Surprisingly, selective use of the diffraction data in the 3.4 A shell improves the overall success rate of the combined-mode classification. An open-source CCP4/CCP4i-compatible version of DIBER is available from the authors' website at http://www.iimcb.gov.pl/diber and is subject to the GNU Public License.