A Corrected Multiscale Stochastic Framework for Combined Interventions in Cellular and Tissue Aging: Coupled Hallmark Dynamics, Global Sensitivity Analysis, and a Falsifiable Translational Roadmap

Aging reflects the progressive breakdown of cellular and tissue homeostasis through interconnected hallmark processes: mitochondrial dysfunction, reactive oxygen species (ROS) accumulation, genomic instability, epigenetic drift, cellular senescence, and chronic low-grade inflammation ("inflammaging") López-Otín et al. (2013, 2023). This manuscript builds on and mathematically corrects an earlier preprint that proposed a stochastic differential equation (SDE) framework for this hallmark network together with simulated NMN/fisetin synergy. In preparing the present formulation we discovered, by attempting to execute the originally reported model verbatim, that its quasi-steady-state ROS closure is dimensionally inconsistent (it evaluates to non-physiological concentrations of order 10–30 M) and that its genomic-instability and epigenetic-drift equations lack any saturating term, so that the coupled system diverges numerically within a few simulated decades. All headline quantitative claims in the original preprint (e.g., a 10^{5}-fold restoration of mitochondrial function, an 85% reduction in post-bifurcation epigenetic noise) were therefore not reproducible from the stated equations and are removed here. We present a corrected, dimensionally consistent, and numerically stable formulation: a dimensionless oxidative-burden index \eta \in [0,1] driven mechanistically by mitochondrial functional decline, and logistic (carrying-capacity-limited) production terms for genomic instability and epigenetic drift, preserving the model's biphasic pre-/post-bifurcation architecture ( T_{b} \approx 45 yr) and its six coupled hallmark state variables. The corrected model is implemented as a fully reproducible, open, vectorized Euler-Maruyama/Milstein simulator (Appendix A) and exercised with n = 2000 antithetic sample paths per intervention arm (Baseline, NMN, Fisetin, Synergy, where "Synergy" denotes simultaneous NMN+fisetin administration under the model's structural assumption of independent multiplicative rate shifts, not a pharmacologically confirmed synergistic interaction - see Section 8). Under the corrected dynamics, combined NMN+fisetin administration from age 45 onward yields, at age 100 relative to the untreated baseline: an 83.5% relative increase in mitochondrial function ( M: 0.102 \pm 0.063 \rightarrow 0.188 \pm 0.113 , Cohen's d = 0.93 ), a 43.3% reduction in genomic instability ( G: 26.6 \pm 1.5 \rightarrow 15.1 \pm 1.7 , d = -7.09 ), a 21.2% reduction in epigenetic drift ( E: 0.493 \pm 0.074 \rightarrow 0.388 \pm 0.074 , d = -1.42 ), a 38.2% reduction in senescence-cell fraction ( S, d = -2.72 ), and a 29.1% reduction in inflammaging ( I, d = -1.72 ; all comparisons Mann-Whitney p < 10^{-16} ). These effect sizes are an order of magnitude more conservative than the original preprint's claims, and we argue they are more credible precisely because they are of comparable magnitude to the effects reported in the underlying NMN and fisetin clinical/preclinical literature Qin et al. (2023); Yousefzadeh et al. (2018) rather than exceeding them by several orders of magnitude. All parameters are literature-informed order-of-magnitude assumptions rather than values fitted to individual-level data, so the model's quantitative outputs should be read as a structured, falsifiable hypothesis rather than a validated clinical prediction (Section 8). We further contribute: (i) a full variance-based global sensitivity analysis (Sobol/Saltelli, N = 4000 base samples, 8 parameters) identifying genomic repair rate \tau_0 ( S_T = 0.83 for G(100) ), epigenetic repair rate \theta_0 ( S_T = 0.67 for E(100) ), and the epigenetic-to-senescence coupling \lambda_s ( S_T = 0.58 for S(100) ) as the dominant drivers of long-horizon outcomes; (ii) a mesoscale spatial reaction-diffusion extension showing that senescent-cell clearance, by converting spatially extended clonal patches into scattered independent foci, reduces Moran's I from 0.973 to 0.718 in simulation; (iii) an explicit scientific and technical risk assessment; and (iv) a falsifiability and experimental-validation roadmap specifying the observations that would confirm or refute each central claim. All code, parameters, and random seeds needed to reproduce every number and figure in this manuscript are given in Appendix A."