ChemBang: Expanding the Chemical Space Around Small Molecules.

Area of Science:

• Computational chemistry
• Drug discovery
• Medicinal chemistry

Background:

• Efficient exploration of chemical space is crucial for generative drug design.
• Existing methods often face challenges in navigating vast chemical landscapes.

Purpose of the Study:

• Introduce ChemBang, a novel computational engine for generative drug design.
• Demonstrate its capability in exploring chemical space and generating drug-like molecules.

Main Methods:

• ChemBang utilizes chemical transformations derived from matched molecular pair analysis of synthesized molecule databases.
• It maps transformations to atomic environments within a three-atom radius.
• Unsupervised chemical evolution is performed by applying these transformations iteratively to seed structures.

Main Results:

• ChemBang successfully generated the Erdafitinib structure from its constituent ring systems.
• The engine guided the generation of drug-like and synthetically accessible molecules by incorporating physicochemical properties and substructural alerts.
• It accurately reproduced 95.3% of 2,809 small-molecule ATC drugs from their fragments, showcasing its pharmaceutical relevance.

Conclusions:

• ChemBang offers an efficient and effective method for exploring chemical space in drug discovery.
• The engine's ability to generate known drugs and guide molecular evolution highlights its potential for designing novel therapeutics.
• This approach advances generative drug design by leveraging chemical transformations and property-guided evolution.