A Spatial Agent-Based Model of AGE-RAGE Feedback in Hepatic Fibrosis Reveals Stage-Dependent Irreversibility Thresholds

Hepatic fibrosis progression involves a well-characterized but computationally unmodeled feedback loop: Advanced Glycation End-products (AGEs) accumulate on permanent collagen via Maillard chemistry, activate the Receptor for Advanced Glycation End-products (RAGE) on hepatic stellate cells (HSCs) and Kupffer cells, drive NF-{kappa}B-mediated HSC activation and anti-apoptotic signaling, and deplete soluble RAGE (sRAGE) through hepatocyte loss -- creating a closed positive feedback loop. To our knowledge, we present the first spatial agent-based model incorporating the complete AGE-RAGE-sRAGE axis in a three-dimensional GPU-accelerated liver tissue simulation. The model produces three key findings: (i) stage-dependent irreversibility thresholds emerge without explicit stage-gating, with resolution declining from [~]68% at F2 to <10% at F4; (ii) sRAGE trajectories diverge at F2-F3: recovering during abstinence from F2 (0.68 [-&gt;] 0.87) but remaining depleted from F3 (0.54), predicting a clinically testable biomarker transition; and (iii) RAGE-driven HSC activation becomes self-sustaining at F3+ independent of exogenous injury, explaining why late-stage fibrosis resists resolution despite removal of the primary insult. No prior computational model -- ODE, PDE, or agent-based -- has formalized the complete RAGE-AGE-sRAGE feedback loop in hepatic fibrosis. The sRAGE divergence prediction is independently testable in clinical cohorts.