From Circuit Descriptions to Testable Mechanism Space in PTSD: A Minimal, Identifiability-Aware Computational Framework for Heterogeneous Threat Inference

Post-traumatic stress disorder (PTSD) is commonly described in terms of frontolimbic circuit alterations and stress-system dysregulation. Yet the central translational problem remains mechanistic: patients with similar symptom severity can differ in why fear persists, why safety fails under stress, and why relapse follows context shifts. Here we synthesize convergent evidence across fear circuitry, noradrenergic and endocrine gain control, and context-dependent learning, and we propose a minimal computational framework that is explicitly designed to be (i) hypothesis-generating rather than confirmatory, and (ii) constrained by identifiability limits in realistic multimodal human datasets. The framework separates fast threat-expression dynamics from slower latent-context inference and learning, and compresses heterogeneity into four composite dimensions: _Control_ (stress-fragile regulation), _Context_ (imprecise or biased context inference), _Gain_ (arousal-amplified expression), and _Recovery_ (feedback and return-to-baseline). We emphasize that “attractor” language is used as an inference-level coarse-graining, not as a directly demonstrated property of human PTSD circuitry. We then derive discriminative, falsifiable predictions and provide a translational mapping that distinguishes plausible causal levers from state-dependent modulators and correlational markers. The goal is not a PTSD biomarker, but an adjudicable mechanism space: a disciplined way to ask which process is dominant in a given person, under which conditions, and with what decision-relevant consequences.