Supervised learning trains a model from labeled examples: each input is paired with a target value or class. In drug discovery, a molecule can be labeled “is toxic” or “is not toxic,” allowing the model to classify new molecules. section 1.2.2
A compound becomes numerical input through molecular descriptors, structural fingerprints, or other features. The model learns an approximate function from these representations to experimentally determined outcomes, rather than observing the full chemical space directly. section 3.3.1
A supervised-learning workflow includes:
- Collect compounds with known labels, such as whether each compound blocks the hERG channel. section 3.1.1
- Represent each compound with features such as molecular weight, bond counts, atom counts, descriptors, or fingerprints. section 1.2.2 section 3.3.1
- Train the model by adjusting weights to reduce prediction error, then freeze those weights during inference on unseen compounds. section 3.3.1
- Check whether predictions generalize beyond the training examples; limited or mismatched data can reduce reliability, such as transferring cell-culture toxicity results to humans. section 1.2.2 section 3.3.1
Predicting hERG blockage
Represent druglike compounds with molecular features and pair each representation with an experimentally determined positive or negative hERG-blocking label. Train a binary classifier, then apply it at scale to previously unseen compounds. section 3.1.1 section 3.3.1
Figure 3.9 Parametric ML model during training. Our ML model learns a function that maps each input data instance to a prediction about a property of interest. The loss and optimization functions measure and aggregate the performance of the model’s predictions. In tandem with a learning routine, the model’s weights are updated such that, in the next iteration of training, the model will reduce its previous error and perform better.

Shows how loss, optimization, and iterative weight updates improve predictions during supervised training. section 3.3.1























