LLM Extraction, Annotation, and Bioinformatics Analysis of Fusion Tuberculosis Vaccines – UROP Symposium

LLM Extraction, Annotation, and Bioinformatics Analysis of Fusion Tuberculosis Vaccines

Kimberly Huang

Research Mentor: Yongqun He
Mentor Department: Not Available, Medicine
Author(s): Kimberly Huang, Leo Yeh, Xingxian Laurel Li, Yongqun Oliver He
Session: Session 5 (2:00 PM – 2:50 PM)
Presentation Type: Poster 10

Abstract

Tuberculosis (TB) remains a major global health challenge, driving continued efforts to develop more effective vaccines. The currently licensed BCG vaccine provides reliable protection in children but shows limited and inconsistent efficacy in adults, highlighting the need for alternative vaccine strategies. In this study, we focused on organizing and curating data related to recent TB vaccine candidates, with an emphasis on subunit fusion protein vaccines. Using resources developed by previous researchers, the VaxLLM vaccine large language model program was applied to extract 3,273 TB vaccine related abstracts from PubMed, along with a spreadsheet containing Mycobacterium tuberculosis vaccine candidates and their corresponding PubMed IDs. These publications were systematically read, analyzed, and curated. Relevant vaccine information was then entered into VIOLIN, a comprehensive vaccine database, following standardized data entry criteria. As a result, detailed data for 10 subunit fusion protein vaccines targeting M. tuberculosis were successfully curated and integrated into the database to be used for future comparison and analysis. An additional 10 fusion protein vaccines are currently in process and are expected to be completed. Overall, this work improved the accessibility, organization and usability of TB vaccine research data. By strengthening information on emerging vaccine strategies beyond BCG, this effort supports future comparative studies and accelerates research to add effective TB vaccines for adult populations.

lsa logoum logo