Structural Rebuilding Confers Superior Long-Term Resilience: A Unified Multi-Mechanism Computational Comparison of Antidepressants in Chronic Stress

Background: Major depressive disorder remains inadequately treated in many patients, with divergent onset speeds and durability across antidepressant classes. Emerging neuroplasticity frameworks emphasize synaptic loss, yet direct computational comparisons of glutamatergic (ketamine-like), monoaminergic (SSRI-like), and GABAergic (neurosteroid-like) mechanisms—particularly under chronic stress—are lacking. Methods: We extended a pruning-plasticity model by applying 95% magnitude-based synaptic elimination to overparameterized feed-forward networks trained on Gaussian cluster classification. From identical pruned states, three interventions were tested across 10 stochastic seeds: ketamine-like gradient-guided regrowth (50% reinstatement) with consolidation; SSRI-like prolonged low-learning-rate refinement with gradual noise reduction; and neurosteroid-like global tonic inhibition (30% damping plus bounded activations). Outcomes included acute recovery, stress resilience (graded internal noise), acute relapse (additional 40% pruning), and longitudinal durability (eight cycles of 10% cumulative pruning with mechanism-specific maintenance). Results: All treatments restored near-ceiling baseline performance, but resilience diverged: ketamine-like networks showed superior extreme-noise tolerance (81.7%) and zero acute/longitudinal relapse across seeds, despite rising sparsity. SSRI-like refinement yielded moderate gains (81.3% combined-stress acutely) but 40% seed relapse longitudinally. Neurosteroid-like inhibition matched early efficacy (97.8%) yet exhibited state-dependence and late-cycle vulnerability (20% seed relapse). Conclusions: These simulations demonstrate mechanistically distinct routes—structural rebuilding (durable), gradual optimization (variable), reversible stabilization (rapid but fragile)—yielding predictive trade-offs in relapse risk. Findings support mechanism-based personalization, prioritizing synaptogenic agents for severe/chronic cases while rationalizing combinations for acute relief and maintenance.