Thermodynamic Detection of Tumorigenesis via Equilibrium Deviations - From Synthetic Dynamics to Translational Oncology

We present a thermodynamically grounded framework for detecting tumorigenesis as a deviation from physiological equilibrium in a geometrically structured state space. Kullback-Leibler divergence quantifies distributional drift from a healthy baseline, while Fisher information geometry provides a Riemannian metric enabling principled adaptive detection thresholds. A structural drift functional over time-resolved interaction graphs captures network-level rewiring alongside distributional changes. Controlled synthetic experiments demonstrate that structural first-passage time anticipates critical transitions earlier than distributional baselines. Single-cell gene expression analyses reveal progressive network reorganization preceding clinical diagnosis. While inferred graphs reflect causal-consistent statistical dependencies rather than definitive causal mechanisms, results indicate that tumorigenesis involves systematic network changes preceding large-scale state deviations. The framework offers a unified, interpretable approach combining thermodynamic principles with structure-aware inference.