Table 2.
Widely used AI techniques in drug discovery
| Category | Task | Method | Representative application | Reference |
|---|---|---|---|---|
| Supervised learning | Regression analysis | MLR | DTI | Talevi et al.29 |
| DT | Adverse drug reactions | Hammann et al.33 | ||
| LR | Drug-drug interaction | Schober and Vetter34 | ||
| Classification | SVM | Compound classification | Maltarollo et al.35 | |
| CNN | Bioactivity prediction | El-Attar et al.36 | ||
| RNN | De novo drug design | Gupta et al.37 | ||
| GAN | Molecule discovery | Blanchard et al.38 | ||
| Unsupervised learning | Clustering | k-means | Drug candidate selection | Shen et al.39 |
| Hierarchical | Molecular scaffold analysis | Manelfi et al.40 | ||
| Dimension reduction | PCA | QSAR | Yoo and Shahlaei41 | |
| t-SNE | Chemical space mapping | Karlov et al.42 |
CNN, convolution neural network; DT, drug target; GAN, generative adversarial network; LR, logistic regression; MLR, multiple linear regression; PCA, principal-component analysis; RNN, recurrent neural network; SVM, support vector machine; t-SNE, T-distributed stochastic neighbor embedding.