Skip to main content

Table 3 The statistics of all generated CoMFA models in order to obtained the best model

From: A combined 3D-QSAR and docking studies for the In-silicoprediction of HIV-protease inhibitors

Charges Model GS q2 SEP C r2 SEE F r pred 2
Ligand-Based Method
Gasteiger Huckel First 0.5 0.77 0.9 6 0.93 0.45 171.85 0.91
  1 0.77 0.9 6 0.93 0.47 173.31 0.91
  1.5 0.74 0.95 6 0.95 0.49 179.4 0.91
  2 0.73 0.92 5 0.92 0.5 181.12 0.91
Best 2 0.71 1.03 6 0.94 0.44 212.63 0.96
AM1BCC First 0.5 0.65 1.12 6 0.9 0.56 117.99 0.87
  1 0.64 1.13 6 0.9 0.56 118.9 0.88
  1.5 0.64 0.96 6 0.94 0.57 116.85 0.88
  2 0.74 1.01 6 0.93 0.42 217.07 0.9
Best 2 0.72 0.94 6 0.92 0.49 173.85 0.9
MMFF94 First 0.5 0.74 0.94 5 0.92 0.51 175.71 0.92
  1 0.74 0.95 5 0.92 0.51 176.35 0.91
  1.5 0.72 0.98 5 0.92 0.51 172.22 0.91
  2 0.78 0.88 5 0.95 0.39 263.64 0.93
Best 2 0.74 0.99 6 0.95 0.42 240.75 0.96
Structure-Based Method
MMFF94 Best 2 0.682 1.09 6 0.93 0.48 178.46 0.93
  1. Where: GS grid spacing, q2: cross validated correlation coefficient, SEP Standard Error of Prediction, C optimal number of Components, r2: non-cross validated correlation coefficient, SEE Standard Error of Estimation, F Fischer test values, r pred 2 : prediction of external test set for validation.