SP-Mind Agent Automates Spatial Proteomics Analysis Workflows

Spatial proteomics is a vital technique for understanding protein expression at a single-cell resolution within tissue structures, offering crucial insights into areas like tumor microenvironments and precision medicine. However, current analysis workflows are often fragmented, demanding significant manual intervention from experts to orchestrate various heterogeneous tools, which limits both the scalability and reproducibility of research. To address these limitations, researchers have developed SP-Mind, an innovative autonomous AI agent. SP-Mind is specifically engineered to streamline and unify the entire spatial proteomics analysis pipeline. It can process raw multiplexed tissue imaging data and proceed all the way to downstream phenotype discovery. The agent is equipped with a comprehensive set of expert-curated biological analysis skills and integrates specialized computational tools. A key feature of SP-Mind is its ability to interpret natural-language queries and translate them into complete, end-to-end analytical workflows without requiring task-specific fine-tuning. To rigorously assess its capabilities, a new benchmark called SP-Bench was introduced, featuring 102 tasks across 18 diverse categories and tissue types. Extensive evaluations on SP-Bench and established downstream tasks demonstrate that SP-Mind achieves state-of-the-art performance, outperforming existing open-source biomedical agent baselines.