Functional Depth Biomarkers Distinguish Lung Squamous Cell Carcinoma from Lung Adenocarcinoma

Lung squamous cell carcinoma (LUSC) and lung adenocarcinoma (LUAD) exhibit fundamentally distinct pathway coordination architectures. We developed a framework integrating pathway activity inference from spatial transcriptomics data, spatial proximity based network construction, and functional depth analysis across 996 TCGA patients. Applying Fraiman-Muniz depth statistics, we generated two representations: Population Referenced Depth quantifies typicality relative to population distributions, while Patient Referenced Depth assesses within-patient network organization. Random forest classification revealed that Patient Referenced Depth marginally outperforms population comparisons, achieving test AUC of 0.768. We focus on bidirectional interaction patterns obtained from spatial interaction networks of pathways. Multi-method feature integration identified three mechanistic frameworks distinguishing subtypes: myeloid orchestrated immune coordination (JAK-STAT $\leftrightarrow$ TNF$\alpha$ dominates LUAD through SPP1$^{+}$ macrophage niches), mutation driven pathway rewiring (TP53 mutations create ecosystem-wide reorganization in LUAD but homogeneous baseline in LUSC), and hypoxia-hormone microenvironment programming (peripheral LUAD tumors coordinate fluctuating hypoxia with angiogenesis while central LUSC tumors integrate chronic hypoxia with death receptor regulation). We identified five novel LUSC-enriched interactions (Androgen $\leftrightarrow$ TRAIL, EGFR $\leftrightarrow$ Estrogen, EGFR $\leftrightarrow$ TNF$\alpha$, Hypoxia $\leftrightarrow$ TRAIL, TGF$\beta$ $\leftrightarrow$ TRAIL) that remain mechanistically uncharacterized, revealing critical knowledge gaps. This framework provides a statistically principled approach for extracting actionable biomarkers from biological networks with applications to precision oncology and therapeutic target identification.