Table 2 Comparison of prediction performance at the residue level on four datasets (YW923, YW150, PDB491 and PDB366) between the proposed RDPred method, the YW method (Yuan and Wang, 2008), and when applying different feature sets and test types (best results are bolded for each dataset and each test type).
| Â | Â | YW923 | YW150 | ||||
|---|---|---|---|---|---|---|---|
Test type | Method | PCC | MAE | MRE (%) | PCC | MAE | MRE (%) |
3 CV | YW method | 0.65 | 0.60 | 18.0 | -- | -- | -- |
| Â | PSSMa | 0.64 | 0.60 (0.049,1.533) | 18.9 (2.4,43.1) | 0.58 | 0.66 (0.057,1.591) | 20.6 (2.7,45.7) |
| Â | PSSM+PSa | 0.66 | 0.59 (0.049,1.502) | 18.7 (2.4,42.6) | 0.61 | 0.65 (0.054,1.581) | 20.4 (2.6,45.5) |
| Â | All featuresa | 0.68 | 0.58 ( 0.048,1.461) | 18.5 (2.3,42.1) | 0.62 | 0.64 (0.056,1.555) | 20.4 (2.7,45.4) |
| Â | RDPred b | 0.67 | 0.56 (0.037,1.510) | 17.0 (1.8,41.4) | 0.63 | 0.61 (0.041,1.557) | 18.1 (2.0,43.2) |
10 CVc | PSSM+PSa | 0.67 | 0.58 (0.048,0.496) | 18.5 (2.3,42.3) | -- | -- | -- |
| Â | RDPred b | 0.67 | 0.56 (0.036,1.501) | 16.9 (1.8,41.1) | -- | -- | -- |
| Â | Â | PDB491 | PDB366 | ||||
Test type | Method | PCC | MAE | MRE (%) | PCC | MAE | MRE (%) |
3 CV | YW method | -- | -- | -- | -- | -- | -- |
| Â | PSSMa | 0.65 | 0.64 (0.054,1.644) | 20.0 (2.6,44.8) | 0.63 | 0.62 (0.053,1.560) | 19.6 (2.6,44.0) |
| Â | PSSM+PSa | 0.66 | 0.64 (0.054,1.625) | 20.0 (2.6,44.6) | 0.63 | 0.61 (0.053,1.539) | 19.6 (2.6,43.8) |
| Â | All featuresa | 0.68 | 0.62 (0.054,1.582) | 19.7 (2.5,43.8) | 0.65 | 0.60 (0.053,1.507) | 19.4 (2.5,43.5) |
| Â | RDPred b | 0.66 | 0.61 (0.041,1.630) | 18.2 (2.0,43.3) | 0.64 | 0.58 (0.040,1.556) | 17.7 (2.0,42.3) |
10 CV | PSSM+PSa | 0.66 | 0.63 (0.053, 1.620) | 19.7 (2.5,44.1) | 0.64 | 0.61 (0.053,1.535) | 19.4 (2.5,43.6) |
| Â | RDPred b | 0.67 | 0.60 (0.040,1.631) | 18.1 (2.0,43.1) | 0.65 | 0.57 (0.039,1.546) | 17.6 (1.9,42.2) |