Project Summary Major depressive disorder (MDD) and associated anxiety disorders are the most prevalent and costly mental illnesses in the United States, with health spending on treatment recently exceeding $71 billion per year. It is now well established that MDD represents a spectrum of disorders, but current drug based approaches to treatment are temporally nonselective, and their efficacy varies highly across individuals. In this proposal, we explore a novel individualized intervention strategy, wherein we aim to prevent and reverse MDD through closed-loop behavioral and neural circuit “tuning”. While some individuals develop MDD as a result of a stressful life event, other individuals appear more resilient to stress-induced depression. Our goal in this proposal is to leverage recent advances in machine learning to identify and detect specific pro-resilient behaviors and patterns of activation in resilient individuals, and then use these data to “steer” susceptible individuals into pro-resilient states. We will accomplish this in two phases. In the first phase, we will test whether modification of behavior alone can generate a pro-resilient state. We will take a novel quantitative approach to behavior analysis, using machine learning to identify specific micro behaviors that are unique to resilient individuals during a chronic social stress. Then, to test whether promoting these behaviors can provide depression-protective effects, we will then use a closed-loop strategy to detect ongoing behavior, and reinforce identified pro-resilient micro behaviors. Second, we will perform circuit-wide calcium recordings in the brain’s subcortical social behavior network and perform unsupervised detection of pro-resilience circuit motifs across the population. We will then use a novel closed-loop read-write strategy to optogentically “tune” the circuit dynamics to mimic these pro-resilient states. We will further explore how these interventions can be accomplished at various time points relative to a stressful life event (before, during, and after) to test whether circuit intervention can potentially provide protective or restorative treatment. These data can potentially be used to develop novel behavior-based therapies for MDD, or to significantly refine the current use of deep-brain stimulation in order to generate pro-resilient states.