AutoDesigner - Core Design, a De Novo Design Algorithm for Chemical Scaffolds: Application to the Design and...

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Traditional scaffold hopping for drug discovery has limitations in chemical space exploration and preliminary structure-activity relationship (SAR) data.
• This can hinder the identification of novel scaffolds with optimal biological activity, selectivity, and properties.

Purpose of the Study:

• To introduce AutoDesigner - Core Design (CoreDesign), a de novo scaffold design algorithm to address limitations in hit identification.
• To systematically explore and refine chemical scaffolds against biological targets, enhancing drug discovery efficiency.

Main Methods:

• CoreDesign utilizes a cloud-integrated, de novo approach to design, evaluate, and optimize billions of molecules in silico.
• Active-learning Free Energy Perturbation (FEP) is employed to assess structural novelty, physicochemical attributes, potency, and selectivity.
• t-Distributed Stochastic Neighbor Embedding (t-SNE) is used for rapid analysis and prioritization of chemical scaffolds.

Main Results:

• Applied to Wee1 inhibitor design, CoreDesign explored over 23 billion molecules, identifying 1,342 novel chemical series.
• A structurally novel 5-5 fused core was rapidly identified, meeting hit-identification criteria.
• Synthesized compounds showed potent Wee1 inhibition and excellent PLK1 selectivity.

Conclusions:

• CoreDesign significantly accelerates the hit-identification stage in drug discovery.
• The algorithm enhances the success probability of drug discovery campaigns by providing high-quality, derisked chemical scaffolds.
• This approach enables faster progression of promising drug candidates through the discovery pipeline.