Skip to main content
. 2019 Jul 30;75(Pt 8):696–717. doi: 10.1107/S2059798319008933

Table 1. Results of the analyses.

CCr and S/N refer to correlation versus resolution and signal to noise, respectively. Acc. and F 1 refer to accuracy and the F 1 score, which assess classifier performance. Feature impact on classifier performance is measured by the ΔF 1 metric and features are ranked (#) by this metric. Full descriptions of all terms are given in the accompanying text.

  Description     LR model Feature impact
Feature/score Model Map Ref CCr S/N Acc. F 1 # ΔF 1
ED features (real-space correlation coefficients)
 CC1 HOH F o − F c Y 0.44 0.61 0.69 0.08 20 −0.04
 CC2 HOH 2F o − F c Y 0.07 1.37 0.80 0.13 21 −0.03
 CC3 SO4 F o − F c Y 0.51 4.70 0.92 0.30 16 −0.06
 CC4 SO4 2F o − F c Y 0.10 9.09 0.97 0.52 12 −0.07
 CC5 SO4 from CC3 [F o − F c]+ N 0.36 0.31 0.66 0.06 10 −0.11
 CC6 SO4 from CC4 [2F o − F c]+ N 0.30 0.58 0.69 0.08 17 −0.05
 CC7 SO4 from CC3 [F o − F c]+ Y 0.38 0.36 0.69 0.07 8 −0.12
 CC8 SO4 from CC4 [2F o − F c]+ Y 0.31 0.64 0.71 0.09 14 −0.06
 CC9 SO4 from CC7 F o − F c N 0.47 3.75 0.90 0.25 13 −0.07
 CC10 SO4 from CC8 2F o − F c N 0.13 8.07 0.97 0.50 7 −0.15
 CC11 HOH from CC1 [F o − F c]+ N −0.20 1.37 0.85 0.16 5 −0.16
 CC12 HOH from CC2 [2F o − F c]+ N 0.01 2.34 0.89 0.22 9 −0.11
 CC13 Av. rotated CC3 F o − F c N 0.52 4.42 0.91 0.28 15 −0.05
 CC14 Av. rotated CC4 2F o − F c N 0.15 9.27 0.97 0.54 11 −0.09
ED features (map values or derived)
 ED1 Standard deviation of CC13 values −0.14 0.22 0.59 0.06 19 −0.04
 ED2 Standard eviation of CC14 values −0.08 0.19 0.48 0.05 18 −0.05
 ED3 Peak F o − F c map volume 0.49 7.01 0.96 0.58 3 −0.74
 ED4 Peak 2F o − F c map volume −0.01 3.17 0.98 0.62 1 −0.83
 ED5 Peak 2F o − F c map value (σ) −0.15 0.88 0.84 0.14 4 −0.38
CC features
 CF1 Distance to closest model atom 0.14 5.24 0.97 0.56 2 −0.81
 CF2 SO4-likeness of local environment 0.03 0.9 0.76 0.10 6 −0.15
Composite scores
 ED score Mapping of 19 ED features 0.15 8.08 0.99 0.92    
 CC score Mapping of 2 CC features −0.11 10.2 0.99 0.86    
 C2 score Pseudo-χ2 for ED/CC scores 0.14 3.50 0.96 0.46    
Total classifier Maximum ED/CC likelihood (2 classes) Acc. = 1.00, F 1 = 0.98