Table 4 The performance of models developed on EGFR100 and EGFR1000 train sets using different PubChem fingerprints and evaluated on validations sets
From: QSAR based model for discriminating EGFR inhibitors and non-inhibitors using Random forest
Trained on | Tested on | Fingerprints count | Sensitivity | Specificity | Accuracy | MCC | ROC |
|---|---|---|---|---|---|---|---|
EGFR100 train set | EGFR100 train set | PubChem 881 | 88.01 | 73.34 | 78.2 | 0.58 | 0.90 |
EGFR100 train set | EGFR100 validation set | PubChem 881 | 91.1 | 68.5 | 76.8 | 0.58 | 0.90 |
EGFR1000 train set | EGFR1000 train set | PubChem 881 | 86.97 | 78.36 | 82.92 | 0.66 | 0.89 |
EGFR1000 train set | EGFR1000 validation set | PubChem 881 | 85.7 | 85.5 | 85.6 | 0.71 | 0.90 |