A Computational and Conceptual Framework for Rational, Multi-Target Design of Sterol 14α-Demethylase-Directed Therapeutics Against Naegleria fowleri: In Silico Modeling, Polypharmacology, and a Falsiable Roadmap Towar

Primary amebic meningoencephalitis (PAM), caused by the thermophilic free-living amoebaNaegleria fowleri, is a fulminant neurotropic infection with a case-fatality rate exceeding 97%and fewer than twenty documented survivors worldwide [5, 6]. No therapeutic regimen testedto date has demonstrated reliable ecacy. This manuscript presents a conceptual, hypothesis-generating computational framework for the rational design of a hypothesized small-moleculeinhibitor of N. fowleri sterol 14α-demethylase (NfCYP51), a genetically and structurally val-idated drug target [1, 10], combined with adjunctive engagement of the mevalonate pathwayand glycolytic enolase. Unlike a conventional drug-discovery report, no synthesis, biochemicalassay, or animal experiment has been performed; every quantitative claim is explicitly labeledas either (i) a literature-anchored fact, reproduced with its original citation and numerical value,or (ii) a hypothesized model input chosen for illustrative systems-pharmacology simulation andclearly agged as such. No model of a non-existent compound can support a categorical ecacyclaim; instead, this framework oers falsiable, testable predictions about the parameter regimea real inhibitor would need to occupy to achieve rapid parasite clearance under a dened dosingschedule. A stochastic pharmacokinetic/pharmacodynamic (PK/PD) model (EulerMaruyamascheme, xed seed 20260802, n = 1000 realizations) is combined with a from-scratch SaltelliJansen Sobol' global sensitivity analysis (base sample size N = 4096, rst-, total-, and second-order indices), a from-scratch Partial Rank Correlation Coecient (PRCC) analysis across fourdosing schedules (q6hq24h, N = 4000 each), and a one-at-a-time (OAT) local robustness scan,to identify which model parameters would most need to be experimentally constrained beforeany translational claim could be evaluated. All code is included in Appendix A, cross-validatedagainst an independent scipy.integrate.solve_ivp reference implementation (maximum ab-solute discrepancy < 0.04 h), and is fully reproducible from the single seed reported. A dedicatedrisk-assessment section and an explicit falsiability/roadmap section are provided, consistentwith Popperian criteria for scientic testability [2