Table 2. Accuracy of inter-residue distance prediction for adenine-binding proteins from the SOIPPA dataset.
Structure dataset | Binding ligand | |||||||||||
ADP | ATP | FAD | NAD | SAH | SAM | |||||||
PCC a | MSE b | PCC a | MSE b | PCC a | MSE b | PCC a | MSE b | PCC a | MSE b | PCC a | MSE b | |
Crystal structures | 0.79 | 6.4 | 0.73 | 8.1 | 0.96 | 2.4 | 0.89 | 4.6 | 0.95 | 1.6 | 0.76 | 7.2 |
High-quality models | 0.75 | 7.3 | 0.72 | 8.6 | 0.92 | 4.6 | 0.86 | 6.1 | 0.94 | 1.9 | 0.88 | 3.6 |
Moderate-quality models | 0.75 | 7.9 | 0.68 | 10.1 | 0.88 | 7.2 | 0.86 | 6.3 | 0.86 | 4.1 | 0.83 | 4.9 |
The Pearson correlation coefficient (PCC) and the mean squared error (MSE) are calculated for the actual pairwise Cα-Cα distances upon the superposition of binding ligands and those predicted by SVR from residue-level scores. The accuracy is reported separately for different binding ligands and target protein conformations including crystal structures, high- and moderate-quality protein models.
Pearson correlation coefficient.
Mean squared error in Å.