The prevalence of risky behavior such as substance use increases during adolescence; however, the neurobiological precursors to adolescent substance use remain unclear. Predictive modeling may complement previous work observing associations with known risk factors or substance use outcomes by developing generalizable models that predict early susceptibility. The aims of the current study were to identify and characterize behavioral and brain models of vulnerability to future substance use. Principal components analysis (PCA) of behavioral risk factors were used together with connectome-based predictive modeling (CPM) during rest and task-based functional imaging to generate predictive models in a large cohort of nine- and ten-year-olds enrolled in the Adolescent Brain & Cognitive Development (ABCD) study (NDA release 2.0.1). Dimensionality reduction (n = 9,437) of behavioral measures associated with substance use identified two latent dimensions that explained the largest amount of variance: risk-seeking (PC1; e.g., curiosity to try substances) and familial factors (PC2; e.g., family history of substance use disorder). Using cross-validated regularized regression in a subset of data (Year 1 Fast Track data; n>1,500), functional connectivity during rest and task conditions (resting-state; monetary incentive delay task; stop signal task; emotional n-back task) significantly predicted individual differences in risk-seeking (PC1) in held-out participants (partial correlations between predicted and observed scores controlling for motion and number of frames [r_p]: 0.07-0.21). By contrast, functional connectivity was a weak predictor of familial risk factors associated with substance use (PC2) (r_p: 0.03-0.06). These results demonstrate a novel approach to understanding substance use vulnerability, which—together with mechanistic perspectives—may inform strategies aimed at early identification of risk for addiction.

在青春期，物质使用等危险行为的发生率增加；然而，青少年物质使用的神经生物学前兆仍不清楚。预测模型可以通过开发预测早期易感性的可推广模型来补充先前观察与已知风险因素或物质使用结果的关联的工作。当前研究的目的是识别和描述易受未来物质使用影响的行为和大脑模型。行为风险因素的主成分分析 (PCA) 与休息期间基于连接组的预测模型 (CPM) 和基于任务的功能成像一起使用，以在参加青少年大脑和认知发展 (ABCD) 研究（NDA 版本 2.0.1）。n  = 9,437) 与物质使用相关的行为测量确定了解释最大量差异的两个潜在维度：寻求风险（PC1；例如，对尝试物质的好奇心）和家族因素（PC2；例如，物质使用障碍的家族史） ）。在数据子集（第 1 年快速跟踪数据；n>1,500）中使用交叉验证的正则化回归，休息和任务条件下的功能连接（休息状态；货币激励延迟任务；停止信号任务；情绪 n-back 任务）显着预测了坚持参与者的风险寻求 (PC1) 的个体差异（控制运动和帧数的预测分数和观察分数之间的部分相关性 [r _p]：0.07-0.21)。相比之下，功能连接性是与物质使用 (PC2) 相关的家族风险因素的弱预测因子 (r _p : 0.03-0.06)。这些结果展示了一种了解物质使用脆弱性的新方法，该方法与机制观点一起可能为旨在早期识别成瘾风险的策略提供信息。