Title: High-dimensional mediation methods for a study of major depressive disorder
Authors: Min Qian - Columbia University (United States) [presenting]
DuBois Bowman - Columbia University (United States)
Bin Cheng - Columbia University (United States)
Abstract: Despite steady progress in the study of major depressive disorder, the pathophysiologic mechanisms underlying both successful antidepressant treatment and the persisting vulnerabilities for offspring remain poorly understood. An integral step toward filling these gaps is to investigate biological variables mediating treatment response and familial MDD risk. However, the set of multimodal neuroimaging candidate mediators may be quite large, with possibly correlated elements, posing challenges conventional methods. While statistical and machine-learning methods for handling large-scale problems continue to emerge, there is a paucity of methods for high-dimensional mediation (HDM) analysis. We present a novel statistical methodology to conduct HDM analysis and apply the new method to identify mediators of the impacts of MDD risk on overall functioning, anxiety, and depression outcomes.