ABSTRACT Excessive alcohol use is a significant and serious public health problem, with 28.6 million US adults meeting diagnostic criteria for current alcohol use disorder (AUD). The stress response is a promising target for understanding vulnerability and possible intervention, as risky drinking patterns are deeply and bidirectionally linked to stress responses. However, key characteristics of the stress response have not yet been leveraged: as studies to date have focused on individual brain regions and static snapshots of brain responses, we cannot capture the predictive potential of the multifaceted stress response, which involves widespread interactions between brain regions and unfolds dynamically over time. Indeed, recent evidence indicates that dynamic whole- brain responses can provide unique insight into stress-related conditions. The goal of this R01 is to leverage advances in machine learning and computational modeling to develop and validate whole-brain biomarkers for stress, and test whether dynamic engagement of these stress networks can predict individual differences in drinking and alcohol-related cognition. Using a combination of secondary analysis (N = 390) and new collection of functional MRI data (N = 100), we will identify and validate stress-predictive neuromarkers, capture dynamic trajectories of stress neuromarkers, and test the consequences of these moment-to-moment dynamics for cognitive mechanisms driving risky drinking. In Aim 1, we will build a connectome-based model that predicts responses to multiple modalities of stress exposure in previously unseen individuals using rigorous cross- validation techniques. We will identify functional connections that predict stress responses in clinically heterogeneous samples as well as those specific to individuals with AUD. In Aim 2, we will create a novel moment-to-moment framework to characterize stress response trajectories in the brain and their alterations in AUD. This framework will enable us to test the hypothesis that, rather than simply having higher or lower engagement of a stress-predictive network, individuals with AUD will show atypical stress network engagement trajectories in response to a stressful event. In Aim 3, we will develop a novel neuroimaging paradigm to quantify the temporal dynamics of brain stress network engagement on memory formation and subsequent drinking. With this design, we will test the hypotheses that: 1) information that is temporally and conceptually congruent with stress will be preferentially encoded; 2) dynamic stress networks will co-fluctuate with a validated neuromarker of attention to facilitate learning; and 3) the timecourse and neural networks by which stress dynamically modulates learning will differ in AUD and predict future drinking. Together, this work will provide new insight into the brain’s stress response, including the ways in which its neural architecture (where), temporal dynamics (when), and consequences for adaptive c...