Document Type
Thesis - Open Access
Award Date
2026
Degree Name
Master of Science (MS)
Department / School
Mathematics and Statistics
First Advisor
Xijin Ge
Abstract
Traditional keyword-based search in genomic databases often fails to capture the latent semantic context of biological studies, leading to a ``Lexical Gap" in research discovery. This thesis presents a high-performance dual-engine discovery system built upon 10,000 studies. By utilizing 1,536-dimensional vector embeddings, with large language models (LLMs) we implement a search engine coupled with a novel Hybrid Boosting layer that integrates semantic intent with factual metadata. Furthermore, we apply a high-dimensional K-Means clustering algorithm for unsupervised topic discovery. Empirical benchmarking demonstrates that the Hybrid Boosting engine achieves a 38.1% gain in precision and a 41.2% increase in Mean Reciprocal Rank (MRR) over pure semantic baselines. Validation via clinical WordClouds and Medical Subject Headings (MeSH) heatmaps confirms that unsupervised clusters align with expert-labeled clinical classifications, providing a scalable framework for cross-dataset exploration in functional genomics.
Publisher
South Dakota State University
Rights
Copyright © 2026 the author
Recommended Citation
Allibah, Lawrence, "Semantic Discovery of Genomic Datasets with AI" (2026). Electronic Theses and Dissertations. 2177.
https://openprairie.sdstate.edu/etd2/2177
Rights Statement
