Deep generative models for ligand-based de novo design applied to multi-parametric optimization

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Multi-parameter optimization (MPO) is a critical challenge in new chemical entity (NCE) drug discovery.
• Deep learning generative models show promise for de novo molecular design, but their application in real-world MPO challenges remains underexplored.

Purpose of the Study:

• To evaluate the efficacy of a ligand-based deep learning generative model for accelerating the discovery of lead compounds meeting multiple biological activity objectives simultaneously.
• To demonstrate the practical benefits of applying AI technology to address MPO in an actual drug discovery project.

Main Methods:

• Quantitative Structure-Activity Relationship (QSAR) models were developed for 11 biological activity objectives, achieving moderate to high performance.
• A deep learning (DL)-based AI de novo design algorithm was employed, coupled with the QSAR models, to generate virtual compounds.
• Synthesized and experimentally validated AI-designed compounds and compared their performance against initial molecules.

Main Results:

• The AI de novo design approach successfully generated 150 virtual compounds predicted to be active across all 11 objectives.
• Synthesized AI-designed compounds met an average of 9.5 objectives (86% success rate), significantly higher than the initial molecules' average of 6.4 objectives (58% success rate).
• One AI-designed molecule exhibited activity across all 11 objectives, and two others were active on 10 objectives.

Conclusions:

• AI-driven de novo design, integrated with QSAR modeling, effectively accelerates the identification of drug leads with improved multi-parameter optimization.
• The AI algorithm identified novel chemical scaffolds and functional groups beneficial for MPO, expanding beyond the initial dataset.
• This study validates the significant value of AI technology in addressing complex MPO challenges within practical drug discovery projects.