Molecular Design with Artificial Intelligence: Progress and Perspectives for Small Molecules.

Area of Science:

• Chemistry
• Artificial Intelligence
• Computational Chemistry

Background:

• Chemistry's progress relies on inverse problem-solving methods.
• Historical advancements include atomistic theory, spectroscopy, and theoretical chemistry simulations.
• Generative artificial intelligence (AI) is poised to drive the next major leap in chemical innovation.

Purpose of the Study:

• To review the development of generative AI in chemistry.
• To discuss various AI models, from traditional to state-of-the-art.
• To explore the potential and challenges of AI in synthesizing novel molecules.

Main Methods:

• Review of traditional generative models like variational autoencoders.
• Analysis of contemporary models including large language models and diffusion models.
• Examination of AI's application in molecule design and synthesis.

Main Results:

• Deep learning techniques, introduced in 2016, have spurred rapid AI development in chemistry.
• AI models offer diverse approaches to chemical problem-solving and discovery.
• Significant potential exists for AI-driven synthesis of new chemical entities.

Conclusions:

• Generative AI represents a transformative force in modern chemistry.
• Understanding AI models is crucial for leveraging their capabilities.
• Addressing the challenges of AI-generated molecule synthesis is key for future advancements.