Structured Expression Dispersion as a Context-Dependent Survival Signal: A Falsification and Mechanism Study in TCGA Glioblastoma (v3.0)

This study investigates whether patient-level gene expression dispersion (coefficient of variation, CV) represents a biologically meaningful survival signal or an emergent statistical property of high-dimensional data. Using TCGA glioblastoma (GBM), the CV metric significantly stratifies survival. However, systematic falsification—including random gene controls, permutation-based structure destruction, cross-cancer testing, network analysis, spectral decomposition, predictive validation, and high-dimensional geometry—demonstrates that the signal is context-dependent, non-generalizable, weakly predictive, and not supported by pathway or network structure. These findings show that survival signals can arise from aggregation effects in high-dimensional systems and emphasize the need for structural validation in omics research.