De novo generation of dual-target compounds using artificial intelligence

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Pharmacology

Background:

• Designing drugs with polypharmacology (multiple targets) is challenging but valuable for drug discovery.
• Rational design of multi-target compounds requires advanced computational methods.

Purpose of the Study:

• To develop and apply AI-based methods for designing novel chemical structures targeting multiple therapeutic proteins.
• To identify compounds interacting with adenosine A2a receptor (ADORA2A) and phosphodiesterase 4D (PDE4D) for bronchial asthma.

Main Methods:

• Employed a fragment-based approach using a genetic algorithm with chemical substructures.
• Utilized a deep learning approach with reinforcement learning and generative adversarial networks.
• Synthesized 10 designed compounds and evaluated their bioactivity against 39 human protein targets.

Main Results:

• Successfully designed chemical structures predicted to interact with ADORA2A and PDE4D.
• Three out of 10 synthesized compounds demonstrated significant binding affinity for both ADORA2A and PDE4D.
• These compounds exhibited high specificity for the intended therapeutic targets.

Conclusions:

• AI-driven methods can effectively design multi-target compounds for drug discovery.
• The developed approach shows promise for identifying novel therapeutics, particularly for complex diseases like bronchial asthma.
• Validated AI-designed compounds with high specificity for key asthma targets, paving the way for further development.