Spatial transcriptomics AI agent charts hPSC-pancreas maturation in vivo

Area of Science:

• Computational biology and bioinformatics
• Genomics and transcriptomics
• Developmental biology and regenerative medicine

Background:

• Spatial transcriptomics generates complex datasets requiring specialized computational and biological expertise for analysis.
• Current analytical methods are often limited to narrow, predefined tasks, hindering comprehensive data interpretation.
• There is a need for advanced tools to automate and democratize the analysis of spatial transcriptomics data.

Purpose of the Study:

• To introduce the Spatial Transcriptomics AI Agent (STAgent), an autonomous multimodal AI agent designed for efficient spatial transcriptomics data analysis.
• To leverage multimodal large language models (LLMs) and specialized computational tools for automated, multi-step analysis.
• To enable autonomous deep research by integrating code generation, visual reasoning, literature retrieval, and report synthesis.

Main Methods:

• Development of STAgent, an AI agent integrating multimodal LLMs with computational tools for automated analysis.
• Application of STAgent to analyze spatial transcriptomics data from human stem cell-derived pancreatic cells (SC-pancreas) during in vivo maturation in mice.
• Utilizing STAgent's capabilities for dynamic code generation, visual interpretation of spatial patterns, and literature-based contextualization.

Main Results:

• STAgent autonomously identified the maturation of endocrine cells into islet-like structures, detailing α- and β-cell arrangement and mesenchymal network expansion.
• Analysis revealed strengthening endocrine-endocrine cell interactions and uncovered spatially resolved biological processes driving maturation.
• STAgent provided mechanistic explanations for spatial patterns, contextualized findings with literature, and generated cohesive insights into human pancreatic development.

Conclusions:

• STAgent establishes a new paradigm in spatial transcriptomics analysis by significantly reducing the expertise barrier and analysis time.
• The agentic approach accelerates biological and biomedical discovery through autonomous, in-depth data interpretation.
• STAgent demonstrates the potential of AI to transform complex biological data analysis and drive scientific innovation.