Integrative multimodal modeling for personalized prediction of Alzheimer's disease progression

Objectives of the proposed research
In this project, we aim to develop an integrative modeling framework to leverage complex multi-modal data for disease modeling, biomarker discovery, and personalized risk prediction in Alzheimer’s disease and related dementias, as well as in aging- and vascular-related processes that contribute to cognitive decline.

Study design
We will analyze data from ADSP, which includes whole-genome sequencing and rich phenotypic information from multiple cohorts encompassing diverse ancestries. Participants with available cognitive assessments, fluid biomarkers, neuropathology, cardiovascular measures, and neuroimaging data will be included. We will use harmonized longitudinal and cross-sectional data across cohorts to enable integrative analyses of genetic, clinical, and imaging measures.

Analysis plan
Genetic variants, fluid biomarkers, neuropathology measures, cardiovascular risk factors, and neuroimaging-derived features will be jointly analyzed to derive latent disease profiles. We will use advanced statistical approaches, for example, multivariate functional mixed models for longitudinal clinical assessments, nonparametric modeling for imaging data, and factor analysis. The resulting profiles will then be evaluated for associations with cognitive decline and symptom onset, informing predictive models for dynamic risk estimation, biomarker validation, and individualized risk stratification across diverse ancestry groups.

This project aims to use large-scale genetic, clinical, and imaging data to better understand Alzheimer’s disease and related dementias. By integrating information on genes, brain imaging, biomarkers, and health measures, we will identify patterns that indicate early disease, discover new determinants of risk, and predict how cognitive health may change over time. Our approach will also help stratify individuals by risk and improve personalized predictions of disease progression, ultimately supporting earlier interventions and more targeted care for diverse populations.