Molecule generation for drug design: A graph learning perspective.

Area of Science:

• Computational chemistry
• Artificial intelligence
• Drug discovery

Background:

• Machine learning, especially graph learning, is revolutionizing scientific fields.
• Molecule design and discovery, particularly in pharmaceuticals, is a key application area.
• De novo drug design leverages advanced computational techniques.

Purpose of the Study:

• To provide a comprehensive overview of state-of-the-art molecule design methods.
• To focus on de novo drug design incorporating deep graph learning.
• To categorize existing methodologies and discuss future research directions.

Main Methods:

• Categorization of molecule design methods into 'all-at-once', 'fragment-based', and 'node-by-node' approaches.
• Review of deep graph learning techniques applied to molecule generation and optimization.
• Identification and discussion of relevant public datasets and evaluation metrics.

Main Results:

• Established a clear categorization framework for de novo molecule design methods.
• Highlighted the role of graph learning in advancing molecule design.
• Presented a consolidated view of datasets and metrics for evaluating molecular generation.

Conclusions:

• Graph learning presents significant opportunities for accelerating drug discovery.
• Further research is needed to address current challenges in automated molecule design.
• Standardized evaluation and datasets are crucial for progress in the field.