Molecular analysis and design using generative artificial intelligence via multi-agent modeling.

Area of Science:

• Computational chemistry and materials science
• Artificial intelligence in scientific discovery
• Molecular engineering and design

Background:

• Advancements in artificial intelligence (AI) offer new tools for complex scientific challenges.
• Large language models (LLMs) are increasingly applied to scientific data analysis and generation.
• Molecular engineering requires efficient methods for target identification and candidate molecule design.

Purpose of the Study:

• To introduce and evaluate the X-LoRA-Gemma LLM for molecular engineering tasks.
• To demonstrate a multiagent AI framework for analyzing, designing, and testing molecular structures.
• To explore the potential of AI in optimizing molecular properties and interactions.

Main Methods:

• Utilized a 7-billion parameter X-LoRA-Gemma LLM with a dual-pass inference strategy.
• Employed a human-AI and AI-AI self-driving multiagent approach for target identification.
• Generated candidate molecules using a multiagent generative design process, including autonomous knowledge extraction.
• Analyzed molecular properties such as dipole moment and polarizability using principal component analysis (PCA).

Main Results:

• The X-LoRA-Gemma LLM successfully identified molecular engineering targets and generated novel candidate molecules.
• Designed molecules exhibited predicted increases in dipole moment and polarizability.
• The multiagent framework demonstrated enhanced problem-solving capabilities across diverse scientific domains.

Conclusions:

• Multiagent generative AI frameworks like X-LoRA-Gemma show significant promise for molecular engineering.
• AI integration can streamline molecular design workflows, leading to innovative solutions.
• Further research is needed to address the challenges and opportunities of AI in molecular science.