Skip to main content
eScholarship
Open Access Publications from the University of California

UCLA

UCLA Electronic Theses and Dissertations bannerUCLA

Statistical and Computational Methods for Electronic Health Records and Genomic Data in Personalized Medicine

Abstract

The integration of genomic and clinical data is pivotal to advancing personalized medicine, where tailored treatment strategies are informed by individual genetic and health profiles. In this dissertation, I present innovative statistical and computational methods designed to enhance the accuracy and applicability of these data in clinical settings. My work addresses key challenges related to precision oncology and genetic risk prediction, focusing on the development of models that account for complex, longitudinal, and error-prone data. First, I propose a probabilistic approach to address genotyping errors inherent in polygenic score (PGS) estimation, particularly when utilizing low-coverage sequencing data. By explicitly modeling uncertainty, this method enhances the reliability and interpretability of genetic risk predictions, improving their practical application in clinical risk stratification. Second, I develop a statistical framework, the Anderson-Gill-based Tumor Dynamics (AG-TD) model, to better capture tumor progression through dynamic clinical measurements, improving our ability to associate somatic mutations with treatment responses in oncology patients. This method demonstrates the value of longitudinal clinical data in complementing traditional survival analyses and supports more informed clinical decision-making. Overall, this dissertation highlights the importance of developing robust, scalable methods to extract actionable insights from noisy, real-life genomic and clinical data, enabling improved precision medicine practices in diverse patient populations.