Advancing precision medicine through machine learning integration of multimodal data for complex diseases Rasika Venkatesh

Format:

Book

Thesis/Dissertation

Author/Creator:

Venkatesh, Rasika, author.

Contributor:

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

Language:

English

Subjects (All):

Bioinformatics.

Genetics.

Physiology.

Biomedical engineering.

0715.

0369.

0800.

0541.

0719.

Local Subjects:

Bioinformatics.

Genetics.

Physiology.

Biomedical engineering.

0715.

0369.

0800.

0541.

0719.

Genre:

Academic theses

Physical Description:

1 online resource (214 pages)

Contained In:

Dissertations Abstracts International 87-12B

Place of Publication:

Ann Arbor : ProQuest Dissertations and Theses, 2026

Language Note:

English

Summary:

Complex diseases such as coronary microvascular disease (CMVD), heart failure (HF), and Alzheimer's disease (AD) arise from the interplay of genetic, molecular, and clinical factors; yet traditional approaches that rely on a single data modality often fail to capture this complexity. As a result, risk prediction remains modest, and opportunities for early diagnosis are often missed. Clinical features and polygenic risk scores (PRS) alone fail to capture the biological processes underlying disease onset and progression, limiting both predictive accuracy and translational impact. To address this gap, we developed and applied machine learning (ML) methods to build integrative risk models (IRMs) that combine genomics, transcriptomics, proteomics, and electronic health records (EHRs) to improve prediction accuracy, refine phenotyping, and reveal novel biological insights. Multiomics association studies of CMVD and HF identified genetic loci and molecular pathways associated with disease susceptibility, demonstrating that genetically-regulated transcriptomic and proteomic information can capture disease heterogeneity and highlight biologically-meaningful underlying mechanisms. We observed similar patterns in AD through association studies and pathway enrichment. We then developed IRMs that leveraged imputed transcriptomic and proteomic features, demonstrating significant improvement of AD risk prediction over PRS alone, illustrating the value of leveraging correlated molecular layers for enhanced risk stratification. In CMVD, we developed multimodal proteomics-based models incorporating real-world clinical imaging, genomic, and proteomic features. These models provided superior risk prediction compared to PRS alone and enabled the identification of previously undefined intermediate patient endotypes. Across these studies and phenotypes, integrative approaches not only facilitated earlier risk prediction, but also identified multimodal biomarkers that provide mechanistic insights into cardiovascular and neurodegenerative disease biology. Together, we establish machine learning-based multimodal integration as a powerful framework for advancing precision medicine by improving predictive accuracy, uncovering disease endotypes, and enabling clinically actionable models. The findings underscore the importance of integrating genomics, transcriptomics, proteomics, and clinical data into unified computational frameworks to capture the interactions driving complex diseases and to support earlier diagnosis and intervention, and targeted prevention in precision medicine

Notes:

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

Advisors: Ritchie, Marylyn D.; Kim, Dokyoon Committee members: Cappola, Thomas P.; Romano, Joseph D.; Setia Verma, Shefali; Tatonetti, Nicholas P.

Ph.D. University of Pennsylvania 2026

Vendor supplied data

Local Notes:

School code: 0175

ISBN:

9798247981077

Access Restriction:

Restricted for use by site license