| COX-2 | Cyclooxygenase-2 |
| RFC | Random Forest Classifier |
| DL | Deep learning |
| GNN | Graph neural network |
| GAT | Graph Attention Network |
| GCN | Graph Convolutional Network |
| MPNN | Message Passing Neural Network |
| TCM | Traditional Chinese Medicine |
| CRC | Colorectal cancer |
| ML | Machine learning |
| MOA | Mechanisms of action |
| ECFP | Extended connectivity fingerprint |
| RFC_ECFP | RFC with ECFPs as model’s input |
| DL_ECFP | DL with ECFPs as model’s input |
| RFC_graph | RFC with molecular graphs as model’s input |
| AP | Average Precision |
| AUC | The highest Area Under the Curve |
| ROC | The Receiver Operating Characteristic |
| TN | True negative |
| FP | False positive |
| DL_ECFP_r | DL model using dimensionality-reduced ECFP |
| DL_graph_r | DL model using dimensionality-reduced molecular graph |
| DL_ECFP_graph | DL model using combination of ECFP and molecular graph |
| RFC_ECFP_r | RFC model using dimensionality-reduced ECFP |
| RFC_graph_r | RFC model using dimensionality-reduced molecular graph |
| RFC_ECFP_graph | RFC model using a combination of ECFP and molecular graph |
| ECE | Expected Calibration Error |