Given concerns about the appropriate diagnosis for children with chronic, severe irritability, we characterized such youth; this work formed the basis for the new diagnosis of mood dysregulation disorder with dysphoria (DMDD) in DSM-5. Since the inception of this project (ZIA MH002786-17), approximately 650 highly irritable youth have enrolled, along with more than 200 youth with ADHD. (Many DMDD patients have ADHD, and youth with ADHD tend to have less irritability than those with DMDD but more than healthy youth; hence they are an appropriate comparison group.) Approximately 75 new patients were recruited this year. Youth with DMDD suffer severe impairment, in terms of medications received, hospitalizations, and standardized measures of function. Irritability is one of the most common psychiatric symptoms in children, but there has been little brain-based research on it, and there are few evidence-based treatments. In 2017-18, we articulated a testable, heuristic, translational model of irritability that continues to guide our research. The model posits that core deficits in pediatric irritability include aberrant responses to frustration and aberrant approach responses to threat. Aberrant responses to frustration implicate reward learning circuitry dysfunction e.g., instrumental learning deficits that prevent adaptation to changing environmental contingencies, or exaggerated prediction error responses to the omission of an expected reward. Since the original formulation of the model, we have updated it to include evidence suggesting that cognitive control deficits (specifically, deficient inhibitory control) may contribute to maladaptive behavior in response to either frustration or threat. Irritability is well-suited for the transdiagnostic, translational approach of the Research Domain Criteria (RDoC). We characterize irritability as a continuous variable, in DMDD and other groups i.e., anxiety disorders, ADHD. We use frustrating tasks during neuroimaging, since a hallmark of irritability is difficulty tolerating frustration. In our work, we have used three different neuroimaging tasks to induce frustration in youth while they undergo fMRI. Our most commonly used frustration task is the affective Posner task, which uses an attentional task and the withholding of expected reward to induce frustration. In published work using this task in 195 youth with DMDD, ADHD, and/or anxiety disorders, and healthy youth, we found that irritability is associated with increased prefrontal and striatal engagement when youth attempt the task after receiving frustrating feedback. In a new sample of 66 youth with DMDD, ADHD, or no illness, we used this task while also acquiring resting state functional connectivity data immediately before and after the task. This design allows us to study behavioral and neural adaptations during and after frustration, and to test associations between these adaptations and irritability. Thus, we study frustration as a dynamic, evolving, whole-brain network process. We used a graph theoretical approach to analyze the data. This approach assumes that the brain is organized into complex subnetworks (modules) of brain regions (nodes). When the clustering of nodes into modules changes, network reconfiguration occurs. Two modules identified at baseline (i.e., pre-task resting-state), an anterior default-mode-temporal-limbic and a fronto-parietal module, contributed most to reconfiguration during and after frustration. Global efficiency indexes the capacity to exchange information among all regions of a network. Only global efficiency of modules present in the post-task resting state predicted self- and observer-rated irritability in previously unseen data. This finding suggests that maladaptive recovery from frustration may play a central role in the pathophysiology of irritability and could guide the development of future interventions. Specifically, we found that self-reported irritability was predicted by global efficiency of a fronto-temporal-limbic module present, while parent-rated irritability was predicted by efficiency of ventral-prefrontal-subcortical and somatomotor-parietal modules. These modules include nodes centrally involved in emotion regulation and reward processing. Importantly, the predictions were specific to irritability; global efficiency was not predicted by anxiety, attention, or hyperactivity ratings. Of note, we also analyzed data acquired pre-pandemic in 50 youth scanned on a second frustration task, the Change task, again with pre- and post-task resting state data. The Change task differs from the affective Posner in the timing of frustration (short blocks of frustration, interspersed randomly with non-frustrating blocks, vs. a long block of non-frustration followed by a long bout of frustration) and in the cognitive task (cognitive flexibility vs. attention orienting). Our work with the Change task yielded a partial replication of the results of the affective Posner work. In both tasks, network reconfiguration was most prominent in the frontal-temporal-limbic and fronto-parietal modules. Also in both tasks, only global efficiency in modules present during the post-task resting state predicted irritability, and the predictive modules included a fronto-temporal-limbic module. However, whereas in the affective Posner task, global efficiency of the fronto-temporal-limbic module in the post-task resting state predicted child ratings of their irritability, this same metric in the Change task predicted parent ratings of the childs irritability. This highlights the importance of research on informant effects in irritability. We have recently completed an analysis of this in 700 youth from our clinic as well as in a publicly available data set and will be preparing this for publication. In addition to the affective Posner and change paradigms, we use a third frustration task. This task, Carnival, was designed to test whether irritability is associated with reinforcement learning deficits at baseline and after frustration. Reinforcement learning is the process by which people learn what behaviors will be rewarded. Deficits in such learning could lead to increased frustration in irritable youth. One component of reinforcement learning is prediction error processing i.e., brain signaling indicating a mismatch between an expected and a received reward. Pre-pandemic, we piloted this task in 34 youth in the scanner and obtained both the expected frustration and expected brain activation associated with prediction error processing. If irritability is associated with instrumental learning deficits, this would have direct treatment implications.