Spatial Transcriptomics and Single-Cell RNA Sequencing in Tumor Heterogeneity: Clinical Biomarker Discovery from Chinese Patient Cohorts

• Stereo-seq sub-cellular resolution (<500 nm) reveals spatial exclusion of CD8+ T-cells by PD-L1+ CAFs in HCC non-responders, with a 1.2-fold enrichment within 50 μm (p=0.003). • In NPC, high LAMP3+ dendritic cell density in stroma correlates with 2.3-year median PFS vs 0.9 years (HR=0.41, p=0.001), suggesting a spatial biomarker for anti-PD-1 combinations. • ESCC neoantigen burden (≥150 mutations/Mb) plus CD8+ T-cell proximity (≤30 μm) predicts durable response with 78% sensitivity and 82% specificity. • Visium FFPE compatibility enables retrospective analysis, but 55 μm resolution misses sub-cellular interactions; Stereo-seq requires fresh-frozen tissue, limiting clinical trial integration. • Computational deconvolution (e.g., MESMER) improves cell-type identification but introduces batch effects; cross-cohort validation is mandatory before clinical adoption.