A1694
Title: Outcome-guided clustering of high-dimensional omics for aging subtypes
Authors: Boyi Hu - Columbia University (United States) [presenting]
Abstract: Clustering is widely used to uncover population heterogeneity in complex diseases. However, most existing approaches analyze high-dimensional omics or clinical phenotypes in isolation and do not jointly integrate these data sources, potentially obscuring biologically and clinically meaningful subtypes. An outcome-guided clustering method called TPClust integrates high-dimensional omics with clinical profiles to identify Aging subtypes. TPClust integrates longitudinal clinical trajectories with high-dimensional omics data to jointly identify clusters by capturing cluster-specific time-varying covariate effects on longitudinal outcomes and performing feature selection in high-dimensional omics data. The theoretical properties of TPClust are investigated and extensive simulations are conducted to evaluate its performance. Applied to Religious Orders Study and Memory and Aging Project (ROSMAP) transcriptomic data and longitudinal cognitive trajectories, TPClust identifies four Aging subtypes, from Resilient (preserved cognition, minimal pathology) to Rapid Decline (severe Alzheimers burden), as well as two intermediate subtypes (Late-Onset Decline and Early Vulnerability) that are not captured by unimodal approaches.
