Supplementary Dataset and Code for: "Topological Entropy of the Tumor Microenvironment Predicts Immune Exclusion and Therapeutic Vulnerability in Oral Squamous Cell Carcinoma"

Background: Spatial heterogeneity in oral squamous cell carcinoma (OSCC) drives immune exclusion and therapeutic resistance through physical barriers. However, current analytical approaches are largely confined to local neighborhood analyses, lacking a unified framework capable of quantifying global tissue topological complexity across modalities. Methods: We integrated 10x Visium spatial transcriptomics (8 OSCC cases), imaging mass cytometry (IMC) spatial proteomics (6,229 single cells), and TCGA-OSCC bulk transcriptomics (206 cases) to develop a Topological Entropy Score (TES) framework based on persistent homology. This framework underwent cross-modal validation at the proteomic level, quantified spatial communication costs for 16 ligand–receptor pairs, employed the PC-stable algorithm to infer conditional independence networks, and projected spatial microdomain signatures onto the TCGA cohort via GSVA to evaluate survival outcomes, immune exclusion mechanisms, and drug sensitivity. Results: TES exhibited a 5.28-fold variation across 8 OSCC samples (permutation P = 0.0171; Hedges' g = 1.280, large effect size). The high-TES group displayed reduced spatial autocorrelation of immune genes and significantly elevated ligand–receptor communication costs (P = 0.007), revealing a "distance disability" phenomenon. In the TCGA cohort, the TES surrogate score was strongly correlated with an immune exclusion gene signature (P = 2.14 × 10⁻⁹) and significantly associated with FAK pathway (rho = 0.416, P < 0.001) and ECM-receptor pathway (rho = 0.594, P < 0.001) activation. The high-TES group exhibited a predicted immunotherapy response rate of only 29.1%, compared to 70.9% in the low-TES group. Furthermore, high-TES tumors showed resistance to standard chemotherapy (cisplatin FC = 1.13, P = 0.010; 5-FU FC = 1.12, P = 0.019) but significant sensitization to dasatinib (FC = 0.72, P < 0.001). Conclusions: TES represents a robust, cross-modal spatial topological biomarker that quantitatively captures FAK/ECM pathway-driven immune exclusion barriers. This study establishes a TES-guided stratified treatment strategy, wherein patients with high TES receive stroma-targeting agents (e.g., dasatinib) followed by sequential immunotherapy to overcome physical barrier-mediated resistance.