ChatSpatial: Schema-Enforced Agentic Orchestration for Reproducible and Cross-Platform Spatial Transcriptomics

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• Spatial transcriptomics enables molecular-level tissue architecture analysis.
• Analyzing spatial transcriptomics data involves numerous computational methods across different ecosystems.
• Current methods require significant researcher effort for tool integration, hindering biological discovery.

Purpose of the Study:

• To present ChatSpatial, a novel platform for unifying spatial transcriptomics analysis.
• To enable researchers to conduct complex analyses through a conversational interface.
• To improve reproducibility and efficiency in spatial transcriptomics research.

Main Methods:

• ChatSpatial utilizes a Large Language Model (LLM) with pre-validated tool schemas.
• Domain expertise is embedded in schema descriptions for context-aware parameter inference.
• The platform is built on the Model Context Protocol (MCP), integrating over 60 methods across 15 categories.
• It supports a unified workflow spanning both Python and R ecosystems.

Main Results:

• ChatSpatial successfully replicated two published studies on ovarian cancer and oral squamous cell carcinoma.
• Validation across seven LLM platforms demonstrated near-deterministic reproducibility at the workflow level.
• The platform facilitates exploratory cross-method analyses and practical triangulation across independent analytical frameworks.

Conclusions:

• ChatSpatial significantly reduces the effort required for spatial transcriptomics data analysis.
• The schema-enforced orchestration ensures high reproducibility for multi-step spatial analyses.
• This platform empowers researchers to accelerate biological discovery by streamlining computational workflows.