Nature子刊：华人学者推出「智能空间组学」技术

Core Viewpoint - The article discusses the revolutionary impact of Smart Spatial Omics (S2-omics) technology in biomedical research, which optimizes region selection for spatial omics experiments, enhancing molecular analysis while preserving tissue structure [2][19]. Group 1: Challenges in Spatial Omics - Spatial omics platforms like Xenium, Visium HD, and CosMx provide single-cell gene expression data but are costly, with sample costs reaching up to $7,000, and have limited tissue capture areas [6]. - Traditional region selection relies heavily on pathologists' subjective experience, leading to labor-intensive processes and variability in results across different laboratories [6][5]. Group 2: S2-omics Overview - S2-omics utilizes AI models to extract features from H&E stained images, simulating molecular heterogeneity to guide experimental design [8]. - The workflow consists of three steps: 1. Feature extraction from tissue images, segmenting them into 8μm×8μm superpixels to capture cellular morphology and tissue architecture [8]. 2. Automatic selection of regions of interest (ROI) based on a scoring system that balances coverage and diversity, allowing user-defined parameters [8]. 3. Prediction of molecular information for unmeasured areas based on selected regions, providing a comprehensive "virtual preview" of the tissue [9][11]. Group 3: Practical Applications - In a gastric cancer sample experiment, S2-omics selected a region covering 7 key tissue clusters, achieving prediction accuracies of 73.8% for cell types and 72.8% for community labels [13]. - In a colon cancer study, S2-omics covered 89.3% of the cells selected by experts while reducing blank areas, thus capturing critical structures more effectively [14]. - For kidney samples, S2-omics optimized the layout of views, successfully capturing glomeruli structures and enhancing data continuity and interpretability [15]. Group 4: Flexibility and Efficiency - S2-omics allows users to specify "positive priors" (e.g., focusing on tumor clusters) or "negative priors" (e.g., ignoring muscle areas), adjusting selection strategies accordingly [16]. - The system can automatically determine the optimal number of regions needed, as demonstrated in breast cancer samples where it identified two 2mm×2mm regions sufficient for capturing heterogeneity [17]. Group 5: Implications for Research - The introduction of S2-omics marks a significant step towards standardization and reproducibility in spatial omics experiments, reducing costs and subjective bias while empowering subsequent experimental designs through virtual predictions [19].