Iterative refinement of a binding pocket model: active computational steering of lead optimization

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Binding affinity prediction models are typically validated using cross-validation or blinded test sets.
• Real-world lead optimization involves iterative selection and synthesis of new molecules.

Purpose of the Study:

• To evaluate computational binding affinity prediction in an iterative, temporal lead optimization setting.
• To assess the impact of selecting compounds based on predicted activity and structural novelty.

Main Methods:

• A series of gyrase inhibitors with known synthetic order was used.
• A predictive model was built and iteratively refined using newly synthesized compound data.
• Compound selection prioritized high predicted activity and quantitative 3D structural novelty.

Main Results:

• Iterative computational selection led to rapid improvements in the activity of selected compounds.
• Incorporating structurally novel compounds identified a more diverse range of active inhibitors.
• The approach demonstrated effective performance in a simulated lead optimization campaign.

Conclusions:

• Iterative computational selection is a viable strategy for accelerating lead optimization.
• Explicitly selecting for structural novelty enhances the discovery of diverse active molecules.
• This method offers a powerful tool for guiding drug discovery efforts.