From genetic risk to molecular trajectories multiomic stratification of alzheimer's disease Erica Harrie Suh

Format:

Book

Thesis/Dissertation

Author/Creator:

Suh, Erica Harrie, author.

Contributor:

University of Pennsylvania. Genomics and Computational Biology., degree granting institution.

Language:

English

Subjects (All):

Bioinformatics.

Neurosciences.

Aging.

Genetics.

0715.

0317.

0369.

0493.

Local Subjects:

Bioinformatics.

Neurosciences.

Aging.

Genetics.

0715.

0317.

0369.

0493.

Genre:

Academic theses

Physical Description:

1 online resource (162 pages)

Contained In:

Dissertations Abstracts International 87-12B

Place of Publication:

Ann Arbor : ProQuest Dissertations and Theses, 2026

Language Note:

English

Summary:

Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by broad biological and clinical heterogeneity. Mild cognitive impairment (MCI) represents a critical stage in the AD spectrum, yet current tools cannot reliably distinguish individuals who will progress to AD from those who will remain stable. Patient stratification at this stage is essential for early intervention and precision medicine. Polygenic risk scores (PRS), the most common genetic stratification tool, are limited by their reliance on clinically defined phenotypes, exclusion of rare variants, and their inability to capture dynamic disease processes. This dissertation addresses these challenges across the genetic and metabolomic layers of AD omics. To base genetic risk estimation on objective biological measures rather than clinical diagnosis alone, we developed an oligogenic risk score that leverages FDG- and AV45-PET neuroimaging biomarkers for gene selection. This approach improves diagnostic classification and risk stratification compared with PRS, while enabling gene-level interpretation. We further assessed the contribution of rare genetic variation, finding that rare variant burden provided limited incremental utility over established common variant risk. Nevertheless, this analysis clarifies the complex multiomic landscape of the disease and establishes a critical baseline for future integrative modeling of non-coding rare variation. Shifting from static genetic risk to dynamic molecular profiling, we introduced a longitudinal lipidomic risk score. This framework combines deep learning-derived temporal lipidomic embeddings with PRS and demonstrates that plasma lipidomic trajectories carry predictive information beyond genetic risk alone. This approach was extended through a time-aware attention model applied to longitudinal plasma lipidomics across four diagnostic trajectory groups, producing a Lipidomic Severity Index (LSI). The LSI detects prodromal AD up to four years before clinical conversion, predicts early conversion risk independently of APOE and PRS, and validates against established cognitive, structural, and molecular biomarkers. Together, these studies demonstrate that AD risk prediction and patient stratification are meaningfully improved by incorporating neuroimaging-informed gene selection, rare genetic variation, and longitudinal metabolomic profiling. These findings advance a more biologically informed approach to precision medicine in AD. Future work will integrate these modalities into unified multiomic frameworks and validate them across diverse populations, facilitating translation into clinical practice

Notes:

Source: Dissertations Abstracts International, Volume: 87-12, Section: B.

Advisors: Kim, Dokyoon Committee members: Shen, Li; Naj, Adam C.; Ritchie, Marylyn D.; Saykin, Andrew J.

Ph.D. University of Pennsylvania 2026

Vendor supplied data

Local Notes:

School code: 0175

ISBN:

9798247982951

Access Restriction:

Restricted for use by site license