Exploring Stress-Induced Neural Circuit Remodeling through Data-Driven Analysis and Artificial Neural Network Simulation

Chronic stress induces behavioral rigidity and neural circuit remodeling, yet the underlying computational mechanisms remain unclear. In this study, we analyze in vivo GCaMP8s recordings from the amygdala-striatal circuits (Giovanniello et al.) and identify a difference in the recovery dynamics of the BLA-DMS and CeA-DMS pathways following acute perturbations. Through data analysis and artificial neural network modeling, we find that structural asymmetry and functional separation–formalized by distinct optimization objectives–are necessary for simulating the experimental results. This suggests that the divergent functions of the BLA-DMS (exploration/precision) and CeA-DMS (stabilization/homeostasis) pathways are inherent, not learned. Under chronic stress, the brain adapts by strengthening the CeA-DMS pathway’s stabilizing role. A robustness test (10 independent seeds) confirmed the statistical reliability of the results (e.g., Recovery Time difference), yet revealed a high Coefficient of Variation in pathway coupling weights (≈ 40%), suggesting the brain preserves structural flexibility even while enforcing rigid functional outcomes. The model replicates the observed outcomes, suggesting that the chronic stress is computationally equivalent to a gain shift favoring the stable CeA-DMS pathway.