Leveraging computational models of neurocognition to improve predictions about individual youths' risk for substance use disorders

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT This K23 proposal seeks to provide an early-career clinical psychologist and neuroscientist (Dr. Alexander Weigard) with the mentorship, training, and resources necessary to launch a career as an independent patient- oriented investigator focused on using advanced computational methods to elucidate etiological mechanisms of substance use disorders (SUDs) and generate meaningful predictions for patients. The candidate will work towards this long-term goal through the completion of a research project focused on assessing whether two advanced computational methods can facilitate the selection of features from neuroscientific data that are relevant for the individualized prediction of SUD risk in youth. Although extant research in developmental neuroscience has identified multiple early risk factors that are associated with development of SUD at the group level, there is currently a dearth of large scale, replicable research in which neurocognitive data are used to make reliable and generalizable predictions of SUD outcomes for individual youth. In the proposed project, the candidate will combine his existing expertise in computational models of cognition with new training in predictive informatics methods to assess whether two advanced computational approaches, a) sequential sampling models (SSMs) of cognition and b) network neuroscience, can be used to extract features from longitudinal neurocognitive data that enhance the prediction of youths’ SUD outcomes. The candidate will conduct extensive analyses with two large data sets (Michigan Longitudinal Study, Adolescent Brain Cognitive Development Study) and collect pilot data with 60 young adults to accomplish the following research aims: 1) Quantify the added benefit of SSM parameters for improving the performance of multivariate SUD prediction models, and 2) Identify the multivariate neural signature of v, an SSM parameter with promising links to substance use, and determine the potential of this signature for predicting a precursor to SUD (substance use initiation in mid-adolescence) in ABCD and differentiating young adults with SUDs in the newly-collected pilot sample. Completion of the following training objectives will ensure that the candidate can both carry out the proposed project and establish himself as an independent investigator who is well-equipped to conduct future projects following from this work: 1) Mastering principles of machine learning model development and testing in longitudinal data sets, 2) building expertise in using multivariate network neuroscience methods for feature selection and prediction, 3) increasing clinical and epidemiological knowledge of SUD risk factors beyond neurocognition, and 4) improving professional skills necessary to become an independent patient-oriented investigator. The proposed K23 aims to take a crucial step towards the development of advanced computational neuroscience methods that may ultimately inform SUD prevention efforts by identifying reliable predictors of individuals’ SUD risk, and to set the candidate up to independently conduct leading edge research in the interest of this larger goal.