Deep Learning in the Optimization Stage
In the optimization stage, the goal is to understand and improve properties of small molecule compounds.
Deep learning can enhance prediction accuracy and efficiency.
Directed Message Passing Neural Networks (D-MPNNs) are deep learning models that learn molecular structures by passing information along directed edges in the molecular graph.

Chemprop, developed at MIT, uses D-MPNNs for predicting molecular properties.
Note: Chemprop is not well-suited for peptides.