Spencer Caldwell
Research Mentor: Jean Nemzek
Mentor Department: Unit for Laboratory Animal Medicine, Medicine
Author(s): Spencer Caldwell, Yongqun He, Jean Nemzek, Leo Yeh
Session: Session 1 (9:00 AM – 9:50 AM)
Presentation Type: Poster 87
Abstract
Sepsis, a dysregulated immune host response identified by organ failure and resulting from prior infection, is the most common cause of in-hospital death and continues to impose massive costs on society, both in terms of human life and financial toll. Researchers have made significant progress in identifying risk factors – such as age and health issues like diabetes – and in analyzing the roles of certain proteins, like TNF, in mitigating sepsis advancement. However, inconsistencies in experimental practices, animal models used for studies, and measurements recorded throughout studies have stymied the rate of this progress. By creating a Python pipeline that utilizes OpenAI’s LLM model GPT-5.1 and incorporates PDF-reading functionality, our laboratory aims to harness the capabilities of LLMs to efficiently extract necessary data from a wide range of sepsis studies from PubMed Central, tabulating it in the format of a simplified SEA-CDM. Developing such a workflow could not only directly address the consistency issues seen across sepsis studies – as normalizing the use of a common data model would reinforce expectations for what kinds of measurements should be taken for each experiment – but could considerably accelerate the pace at which sepsis research is conducted. It could additionally help researchers identify hidden relationships between variables contributing to sepsis progression, given SEA-CDM’s reliance on ontologies.
