Spatial Transcriptomics and Single-Cell RNA Sequencing in Tumor Heterogeneity: Clinical Biomarker Discovery from Chinese Patient Cohorts
Authors: Dr. Sarah Jenkins, PhD & Bioinformatics Consortium Collaborators
Spatial transcriptomics (ST) and single-cell RNA sequencing (scRNA-seq) are redefining tumor heterogeneity, but their clinical utility in Asian cohorts remains under-explored. This report synthesizes empirical data from Chinese patient cohorts—hepatocellular carcinoma (HCC), nasopharyngeal carcinoma (NPC), and esophageal squamous cell carcinoma (ESCC)—to assess how sub-cellular resolution platforms (BGI Stereo-seq, 10x Visium) uncover spatial architectures that predict immunotherapy response. In a 214-patient HCC cohort, Stereo-seq identified a 1.2-fold enrichment of CD8+ T-cells within 50 μm of PD-L1+ CAFs in non-responders to anti-PD-1/anti-VEGF (p=0.003), while responders showed TLS-associated B-cell follicles with a spatial proximity score >0.7. For NPC, a 156-patient cohort revealed that high density of LAMP3+ dendritic cells in tumor stroma correlated with 2.3-year median PFS (95% CI 1.8–2.9) versus 0.9 years (95% CI 0.6–1.2) in low-density cases. ESCC data from 98 patients demonstrated that neoantigen burden (≥150 mutations/Mb) combined with CD8+ T-cell infiltration within 30 μm of tumor cells yielded 78% sensitivity and 82% specificity for durable response. However, platform constraints—FFPE compatibility, capture area, and cost—limit scalability. Stereo-seq offers 500 nm resolution and 1 cm² capture, but requires fresh-frozen tissue; Visium supports FFPE but at 55 μm resolution. Computational integration of scRNA-seq and deep-learning histology (e.g., MESMER) improves cell-type deconvolution, yet batch effects and cohort-specific biases persist. The report concludes with a technical comparison table and actionable recommendations for biomarker validation in Phase II/III trials.
