Project Summary The overarching goal of this project is to develop innovative, robust and plausible analytical methods to uncover individualized biomarker trajectories that interrelate with Alzheimer's onset during asymptomatic stage, dissect their associated genetic bases, and dynamically predict the overall disease risk composited with quality of life through massive and time-varying health and biomedical profiles. Alzheimer's disease (AD) is incurable, and its soaring prevalence has induced a global crisis on health and finances. Recent research reveals that AD is a continuum with pathological changes launched years before the emergence of clinical symptoms. The ongoing biomarker research plays a dominate role in tracking disease evolution and predicting AD-related outcomes, and the more accessible electronic health records (EHRs) nowadays further provide an untapped resource for a prompt management of disease progression. However, existing disease dynamics and predictive studies suffer with 1) ignoring the interplay between biomarker dynamics and disease hallmarks, 2) inadequate power under sparse and irregular measurements, 3) failure to handle time-dependent EHRs with subject-specific landmarks, and 4) oversight on predicting risk profiles accounting for patients’ quality of life. To address these barriers, the current project proposes the following aims: Aim 1) to construct AD biomarker trajectories interrelated with disease onset during asymptomatic stage and dissect associated genetic risk profiles; Aim 2) to build dynamic risk prediction and quality of life assessment tools for AD-related events integrating electronic health records, brain imaging traits and neuropsychological metrics; Aim 3) to perform systematic evaluation for the proposed methods through extensive simulations and real data analyses, and develop user-friendly analytical pipelines for the proposed methods. This project is innovative in multiple aspects for and beyond AD medical and biomedical research including but not limited to a) establish multi-domain biomarker trajectories interacted with disease onset, b) consider age and time-to-event indices for marker dynamics as well as flexible and knowledge- driven shapes, c) uncover relevant genetic underpinnings d) account for sampling bias due to delayed entry, e) develop dynamic prediction with subject-specific landmarks, f) predict risk profiles accounting for the life quality, g) develop efficient and user-friendly pipelines for our products. We will implement the proposed paradigms on three large-scale AD cohort studies containing multi-domain repeatedly measured biomedical and clinical data, with one of them linked with a massive EHR dataset of over 2.5 million patients. A successful completion of this project will pave unique ways to achieve early detection, intervention and management for AD. By contributing on laying the groundwork for proactive disease modeling based on multi-domain data sources, we anticipate t...