Using Tsetlin Machine for decoding, visualization and minimization of local immune fingerprints in peritoneal dialysis infections

Gorbenko, Anatoliy, Tarasyuk, Olga, Zhang, Jingjing, Shafik, Rishad, Yakovlev, Alex, Eberl, Matthias ORCID: https://orcid.org/0000-0002-9390-5348 and Topley, Nicholas 2025. Using Tsetlin Machine for decoding, visualization and minimization of local immune fingerprints in peritoneal dialysis infections. Presented at: 2025 International Symposium on the Tsetlin Machine (ISTM), Rome, Italy, 8-10 October 2025. 2025 International Symposium on the Tsetlin Machine (ISTM). IEEE, pp. 132-139. 10.1109/istm67926.2025.00027

Full text not available from this repository.

Abstract

The immune system's primary functions include recognizing invading pathogens, controlling infections, and restoring tissue integrity. However, definitive evidence showing that an individual's local immune system can differentiate between bacterial pathogens to mount specific responses remains limited. In this study, we applied a machine learning approach to characterize immune responses in 82 peritoneal dialysis patients presenting with acute peritonitis. Immune profiles were obtained from peritoneal effluents on the day of infection onset, analyzing a comprehensive array of cellular and soluble markers, including local immune cell populations, inflammatory and regulatory cytokines, chemokines, and tissue damage-associated factors. Utilizing the Tsetlin Machine, a novel logic-based machine learning algorithm, we identified pathogen-specific immune fingerprints for different bacterial groups, each characterized by distinct biomarker profiles. Unlike traditional black-box models, the Tsetlin Machine expresses immune responses as transparent logical rules, enabling visual interpretation and supporting timely, informed antibiotic selection based on a patient's immune profile well before culture results become available. We also present during-and post-training techniques for feature and clause minimization to produce more concise rules, improve interpretability, and reduce computational cost. This capacity for transparent decision-making illustrates the potential of the Tsetlin Machine to analyze complex biomedical datasets and improve patient outcomes by delivering clear and actionable insights.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Publication
Status:	Published
Schools:	Schools > Medicine
Publisher:	IEEE
ISBN:	9798331569235
Last Modified:	07 Jan 2026 10:30
URI:	https://orca.cardiff.ac.uk/id/eprint/183636

Actions (repository staff only)

Edit Item