Answers are grounded in this paper and cite their sources.
Background: Artificial intelligence (AI) can identify neuroimaging signatures associated with individual differences in substance use disorders (SUDs), but predictive performance does not establish biological or clinical validity. Objective: This narrative review ex-amines how clinical prediction can be connected to preclinical mechanistic investigation and cross-species translation. Approach: We thematically synthesized representative clinical and preclinical studies identified primarily through PubMed and reference-list searches, focusing on AI-based biomarker discovery, experimentally testable mechanisms, and translational integration. Main findings: Clinical studies illustrate applications in classification, craving and abstinence prediction, and treatment outcomes, whereas preclinical studies offer complementary behavioral, circuit-level, and molecular evidence and opportunities for causal testing. Cross-species network and phenotype alignment may help connect these approaches, although direct validation of human-derived signatures in experimental models remains limited. Conclusions: A bidirectional framework linking prediction, mechanistic testing, and clinical re-evaluation may improve the interpretability and translational potential of AI-derived signatures.
No takes yet. Share an insight, caveat, or question.
Duan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: