Longitudinal and Predictive Modeling to Relate Self-Evaluation and Depression in Adolescent Girls

PROJECT SUMMARY Depression–which has increased rates of onset during adolescence–is a leading cause of disability worldwide, particularly among adolescent girls. Self-development is a key task of adolescence, but individuals with depression exhibit biases in self-perception. Of note, both self-evaluation and depression have neural

foundations that overlap, including in the ventromedial prefrontal cortex (vmPFC). However, how self- evaluative processes and depression relate over time is still unclear. Longitudinal designs that track the interrelations between neural and behavioral indices of self-evaluative processing and depression are needed

to understand the development of this disorder. Additionally, prospective prediction of depression prior to depression onset–which is important to minimize disease burden and associated healthcare costs–has thus far proved challenging. Longitudinal datasets capturing individuals prior to depression onset, combined with novel

analytic approaches, may prove fruitful for tracking transactional, longitudinal relations between self-evaluation and depression and for prospectively predicting depression. This project will achieve these objectives using three waves of data from a longitudinal study of adolescent females (initial N = 174, initial ages 10-13, 18

months between waves; at least 31% of sample has a clinical depressive disorder at waves 2 or 3). The specific aims are: 1) characterize the cross-sectional and prospective associations between neural and behavioral indices of self-evaluation and depression across three waves of data; 2a) predict depression

diagnosis prospectively via a machine learning classifier approach built on self-evaluative behavior; and 2b) predict depression diagnosis prospectively via multivariate pattern analyses of brain activity during self- evaluation. Completion of these aims will result in a body of knowledge that tracks developmental interrelations

between self-evaluation and depression and predicts depression onset using a clinically meaningful target with high potential for translation. This project has four training goals to help me achieve the project objective and to prepare me for a career as an independent clinical developmental scientist: develop expertise in 1) open and

reproducible science; 2) modeling longitudinal transactional relations between behavior, brain development, and mental health; 3) predictive analysis techniques using machine-learning methods and multivariate approaches; and 4) inclusive research, professional development, and science communication. The University

of Oregon’s state-of-the-art Center for Translational Neuroscience provides the ideal environment for completion of these training goals. Dr. Jennifer Pfeifer will be the primary training mentor overseeing the project due to her extensive knowledge of adolescent self-evaluative development and developmental

neuroimaging. Dr. Nicholas Allen will serve as the clinical expert providing insights on adolescent social development and depression. Dr. Robert Chavez, an expert in multivariate neuroimaging, will mentor me on training and research aims related to multivariate predictive modeling.