Machine learning discovery of patterns of self regulation in drug addiction and I

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): This application is written in response to the RFA on Basic Research on Self-Regulation (RFA-AG-11-010). Disorders where poor self-regulation is a prominent feature involve great harm and pose a serious concern to public health, yet little is known about their underlying neurobiological mechanisms. A series of brain-behavior studies at our laboratory brought forth an empirically based theoretical model of human drug addiction, characterized by Impaired Response Inhibition (RI) and Salience Attribution (SA) (hence, I-RISA). The model posits that addiction involves assigning a lower importance (salience) to non-drug emotional stimuli (while over-valuing drug-related stimuli) with a concomitant compromise in inhibiting disadvantageous responses (e.g., compulsive drug-taking). Neuroimaging mapped these I-RISA components onto dysfunctional striatal- prefrontal cortical circuitry demonstrating the diathesis for impaired self-regulation in this disorder. In the current proposal we will test the I-RISA model in another externalizing psychopathology characterized by impaired self-regulation. Specifically, we will target Intermittent Explosive Disorder (IED), that similarly to addiction, is a chronic and relapsing disorder, featuring a skewed SA and disrupted RI (individuals with IED perceive provocation where none may have been intended, reacting with disproportionate anger that intermittently culminates in assault behavior and damage to property). In both disorders, we will target sensitive brain-behavior measures of self-regulation, using the theory-informed multidimensional datasets to develop novel computer science algorithms to conduct group classification (distinguishing between cocaine addicted individuals, IED, and healthy controls). This project represents a major departure from the current functional neuroimaging and mental health research paradigms in its focus on: (1) abstract reinforcement with money (a universal secondary reinforcer that acquires its value and uniquely impacts human emotional learning and self- control through social communication); (2) positive but also negative reinforcement (going beyond the reward principle to study compromised sensitivity to punishment and adversity); using both to predict (3) self- regulation during neuroimaging (going beyond self-report as further bolstered by psychophysiological measures); and (4) the multimodal platform to automatically perform group classification (and other machine- learning techniques, e.g., multitask) such that the common neurobehavioral signatures (but also discriminative properties) of impaired self-regulation can be identified, a prototype to be generalized to other disorders of self- regulation. The size of the potentially impacted community is of significant proportions: according to current estimates, up to 20% of the adult population in the U.S. suffers from psychiatric symptoms that impair ability to exercise self-regulation. Bringing forth significant gains toward the goal of liberating patients from the cycle of relapsing behaviors (drug use or assault behaviors) that bear catastrophic consequences to the patients themselves and with devastating costs to the broader society, this tool is estimated to be of great value.