Deep Generative Models in De Novo Drug Molecule Generation

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Traditional drug discovery is labor-intensive and costly, relying on experimental lead molecule optimization.
• Artificial intelligence (AI) shows significant promise for efficient and rapid drug-like molecule generation.
• Deep generative models are at the forefront of AI-driven *de novo* molecule design.

Purpose of the Study:

• To review recent advancements in molecule generation using deep generative models.
• To provide a comprehensive comparison of state-of-the-art AI frameworks for molecule generation.
• To identify challenges and suggest future research directions in AI-based drug discovery.

Main Methods:

• Review of literature on deep generative models for molecule generation.
• Focus on molecule representations, public databases, and data processing tools.
• Comparative analysis of leading AI-based molecule generation frameworks and design strategies.

Main Results:

• Deep generative models excel at *de novo* generation of drug-like molecules with specific properties.
• A comprehensive comparison of current AI models and molecular design strategies is presented.
• Key research gaps include database limitations, lack of 3D information, and imprecise evaluation metrics.

Conclusions:

• AI, particularly deep generative models, offers a powerful and efficient approach to novel drug discovery.
• Addressing current limitations in data, representation, and evaluation is crucial for advancing AI in molecular design.
• Future directions involve improving databases, incorporating 3D structural information, and developing precise metrics for AI-driven drug discovery.