SpatialAgent: An Autonomous AI Agent for Spatial Biology

SpatialAgent, an autonomous AI, outperforms experts and baselines in spatial biology tasks across multiple tissues, including designing gene panels and analyzing millions of cells.

• Why it matters: Understanding tissue organization at the molecular level is crucial for biomedical advances, but current workflows are labor-intensive and slow, limiting discovery and application.
• What they did: The team developed SpatialAgent, which integrates large language models with a Plan-Act-Conclude architecture, multimodal interpretation, and verification modules to support the entire discovery process, from gene-panel design to hypothesis generation, tested on diverse datasets.
• The result: SpatialAgent not only surpasses computational baselines but also matches or exceeds expert performance, enabling rapid, autonomous insights into tissue structure, cell interactions, and disease states, exemplified by a successful prostate cancer study.