Modeling Antidepressant-Induced Manic Switch and Longitudinal Relapse: A Unified Pruning Framework Highlights Glutamatergics' Disease-Modifying Potential

Antidepressants that target different brain pathways—glutamatergic (ketamine-like), monoaminergic (SSRI-like), and GABAergic (neurosteroid-like)—vary in how quickly they work, how long effects last, and the risk of triggering mania, especially in people with bipolar disorder. A computer model simulating depression showed that while all three restored normal performance initially, the ketamine-like approach rebuilt neural connections, leading to better resilience under extreme stress and no manic relapse after stopping treatment. The neurosteroid-like approach worked rapidly but caused relapse in 88.3% of cases when discontinued. The SSRI-like method was slowest, showed the highest risk of mania, and led to 95.0% relapse after cessation. These results suggest that selecting antidepressants based on their mechanism could improve safety and long-term outcomes.

Study at a glance

Abstract

Background: Major depressive disorder involves impaired neural plasticity, yet antidepressants targeting glutamatergic (ketamine), monoaminergic (SSRIs), and GABAergic (neurosteroids) pathways differ markedly in onset speed, durability, and risk of treatment-emergent mania—particularly in bipolar contexts. Clinical comparisons are confounded by heterogeneity; computational models enable controlled mechanistic dissection, but few integrate manic liability and post-discontinuation stability across classes. Methods: We extended a magnitude-based pruning model (95% sparsity) of depression in feed-forward networks classifying Gaussian blobs. From identical pruned baselines, three interventions were simulated: ketamine-like gradient-guided synaptic regrowth (50% reinstatement) with consolidation; SSRI-like prolonged low-rate refinement with tapering noise and escalating excitability gain; neurosteroid-like global tonic inhibition (0.7× damping, tanh activations, reduced gain). Efficacy assessed classification accuracy under clean, noisy, and combined stress; resilience via graded noise tolerance; acute relapse after further pruning; manic risk through biased positive perturbation and activation magnitude. Longitudinal relapse modeled chronic maintenance (with mood stabilizer protection) followed by discontinuation, using treatment-specific lingering decay rates. Metrics averaged across 10 seeds. Results: All treatments restored near-ceiling performance acutely, but ketamine-like regrowth yielded superior extreme-stress resilience (76.8%) and zero post-discontinuation manic relapse, reducing sparsity to 47.5%. Neurosteroid-like modulation matched rapid recovery (97.6%) but showed state-dependence and 88.3% relapse probability off-drug. SSRI-like refinement lagged in resilience (49.9% extreme) with highest manic proxies (biased accuracy 47.2%, gain 1.60) and 95.0% relapse post-cessation. Longer maintenance conferred negligible added protection for reversible mechanisms. Conclusions: Antidepressants operate via divergent plasticity routes—durable structural rebuilding (ketamine-like, low long-term risk), rapid reversible stabilization (neurosteroid-like), and vulnerable gradual optimization (SSRI-like)—reproducing clinical trade-offs in speed, persistence, and bipolar safety. These findings support mechanism-guided selection, positioning synaptogenic agents for recurrent or high-risk cases pursuing remission beyond treatment.