SpaCellAgent Automates Single-Cell Trajectory Analysis with LLM Multi-Agents.

The paper introduces SpaCellAgent, an innovative autonomous multi-agent framework powered by large language models (LLMs), specifically designed for spatial and single-cell transcriptomics. These fields are crucial for understanding cellular dynamics and reconstructing cell developmental paths through trajectory inference (TI). However, existing TI methods often demand extensive manual intervention and expertise in diverse tools, creating a significant barrier for researchers. SpaCellAgent aims to bridge this gap by automating the entire spatiotemporal analysis pipeline and narrative generation. Its architecture comprises a multi-agent system for strategic workflow planning, a dynamic tool-orchestration engine for adaptive algorithm selection, and a self-evolution module that continuously refines performance through feedback mechanisms. The framework was rigorously evaluated on six diverse datasets, covering complex temporal developmental trajectories, various sequencing platforms, and spatially-resolved tissue architectures. SpaCellAgent consistently demonstrated over 40% improvement in analytical efficiency while achieving performance levels comparable to expert analysis. By translating natural language specifications into optimized analytical workflows and fully automating the process, SpaCellAgent democratizes advanced spatiotemporal modeling and establishes a scalable, agent-driven paradigm for computational biology.