Drug Development Applications

1 Target Identification

Sensitivity analysis (Equation sensitivity) identifies the model parameters whose modification produces the greatest improvement in the symptom generation functions (Equation symptom mapping). Parameters with high sensitivity indices are candidate drug targets. The integrated model predicts that the highest-sensitivity parameters are:

\(\alpha_\text{CI}\) (Complex I activity): directly limits ATP production capacity
\(k_\text{exh}\) (immune cell exhaustion rate): controls the size of the exhausted immune cell compartment
\(n_F\) (cortisol feedback sensitivity): determines HPA axis set point
\(P_0\) (baseline BBB permeability): gates CNS exposure to peripheral inflammation

These targets correspond to distinct pharmacological classes: mitochondrial protectants, immune checkpoint modulators, glucocorticoid receptor modulators, and BBB-stabilizing agents. Some of these are already under investigation in ME/CFS (e.g., mitochondrial support supplements), while others (checkpoint modulators, BBB agents) represent novel directions suggested by the model.

2 Virtual Clinical Trials

The model enables in silico clinical trials: simulating the response of a virtual patient population to a candidate treatment. A virtual population is generated by sampling parameter vectors from distributions representing the known heterogeneity of ME/CFS. Each virtual patient is simulated under treatment and control conditions, generating predicted effect sizes, responder rates, and optimal trial designs (duration, endpoints, enrichment criteria).

Virtual trials are particularly valuable for ME/CFS because:

Real clinical trials are expensive and slow to recruit
The heterogeneity of ME/CFS means that average treatment effects may be small even when a subgroup responds dramatically—virtual trials can identify the responding subgroup and design enrichment strategies
Ethical constraints limit the types of perturbation studies possible in patients—the model permits simulation of interventions that would be impractical to test in humans

Open Question: Digital Twins for ME/CFS

Could a patient-specific computational model—a “digital twin”—be maintained and updated throughout the course of illness, receiving real-time data from wearable sensors and periodic biomarker panels? Such a system would continuously refine its parameter estimates and predictions, enabling adaptive pacing, early detection of relapses, and prospective treatment optimization. The technical requirements are substantial (real-time model fitting, sensor integration, uncertainty quantification), but analogous approaches are under development for cardiac care and diabetes management. The feasibility for ME/CFS depends on whether the biological signals accessible through non-invasive monitoring contain sufficient information to constrain the model’s critical parameters.