De novo molecular design and generative models.

Area of Science:

• Medicinal Chemistry
• Computational Drug Discovery
• Artificial Intelligence in Pharmacology

Background:

• Molecular design is crucial for developing new therapeutics.
• Computational methods for de novo molecular design have evolved over 30 years.
• Recent advances in machine learning (ML) and artificial intelligence (AI) have significantly impacted drug discovery practices.

Purpose of the Study:

• To review learnings from computational de novo molecular design approaches.
• To categorize de novo design methods based on molecular representation (atom-based, fragment-based, reaction-based).
• To highlight challenges and future opportunities in computational drug design.

Main Methods:

• Categorization of de novo molecular design strategies based on representation granularity.
• Review of computational approaches including atom-based, fragment-based, and reaction-based paradigms.
• Discussion of benchmark importance and practical implementation challenges.

Main Results:

• De novo molecular design methods can be classified by their representation level.
• Practical application of these computational tools faces several challenges.
• The field has gained valuable experience with ML/AI-driven approaches.

Conclusions:

• Computational de novo design, especially with ML/AI, offers significant potential for therapeutic progress.
• Standardized benchmarks and addressing practical challenges are key for future success.
• Further exploration of novel approaches and overcoming existing hurdles will drive innovation in drug discovery.