Individualized Modeling of Neurodevelopmental Network Dynamics Predicting Cannabis Risk and Escalation in Adolescence - PROJECT SUMMARY/ABSTRACT Cannabis is the most commonly used illicit substance among U.S. adolescents, with initiation rising sharply between ages 14–16. Early use increases risk for later substance-use disorders, affective dysregulation, and cognitive impairment, yet the neural mechanisms that predispose some youth to early initiation remain poorly understood. Identifying these mechanisms during sensitive periods of brain maturation is critical for prevention. This F32 fellowship leverages large-scale, longitudinal neuroimaging data from the Adolescent Brain Cognitive Development (ABCD) Study (N ≈ 6,133; ages 9–16) and biochemically verified cannabis exposure (11,865 hair samples; THC-COOH positivity ~7% at mid-adolescence, rising to ~12% by ages 15–16) to discover reproducible, connectome-based biomarkers of cannabis-use vulnerability. The proposed study is among the first to apply fully data-driven, connectome-wide predictive modeling of resting-state fMRI to forecast cannabis initiation and escalation, without imposing a priori regional hypotheses. Aim 1 uses dimensionality reduction and connectome-based predictive modeling (CPM) to identify multivariate connectivity signatures that predict subsequent cannabis initiation. Aim 2 models longitudinal change in these signatures from ages 9–18 to determine how early risk profiles evolve across adolescence and relate to escalation trajectories derived from combined biochemical (THC-COOH) and self-report indices. Aim 3 tests reproducibility, cross-cohort generalization, and uncertainty calibration in the Human Connectome Project–Young Adult (HCP–YA) dataset, benchmarking performance before and after harmonization (e.g., ComBat/ComBat-GAM) and mapping predictive features onto reward, salience, default-mode, and control networks. This multidisciplinary, data- driven program of research integrates developmental neuroscience, computational modeling, biostatistics, and clinical addiction science under the mentorship of Drs. Natania Crane, Olusola Ajilore, Krista Lisdahl, Monica Rosenberg, Runa Bhaumik, Adam Leventhal, and Damien Fair. Structured mentorship (biweekly primary- mentor meetings, monthly computational and developmental co-mentor sessions, and quarterly team reviews) will address key training gaps in developmental neuroscience, machine learning, and longitudinal harmonization while supporting preparation for a K-level award. Situated within the collaborative environment of the University of Illinois Chicago, this innovative extension of large-scale, data-driven connectomics directly advances NIDA's strategic priority to “leverage advances in data science and neurodevelopmental research to identify risk and resilience factors for substance use.” By producing the first large-scale, fully inductive, data- driven, and reproducible connectome-based biomarker of cannabis-use vulnerability, this project establishes a mechanistically interpretable framework for early identification and prevention. This directly fulfills NIDA's strategic goal to apply neurodevelopmental and data-science approaches to identify emerging risk before substance use begins.
