Prediction of the infecting organism in peritoneal dialysis patients with acute peritonitis using interpretable Tsetlin Machines.

Motivation: The analysis of complex biomedical datasets is becoming central to understanding disease mechanisms, aiding risk stratification and guiding patient management. However, the utility of computational methods is often constrained by their lack of interpretability, which is particularly relevant in clinically critical areas where rapid initiation of targeted therapies is key.

Results: To define diagnostically relevant immune signatures in peritoneal dialysis patients presenting with acute peritonitis, we analysed a comprehensive array of cellular and soluble parameters in cloudy peritoneal effluents. Utilizing Tsetlin Machines, a logic-based machine learning approach, we identified pathogen-specific immune fingerprints for different bacterial groups, each characterized by unique biomarker combinations. Unlike traditional 'black box' machine learning models, Tsetlin Machines identified clear, logical rules in the dataset that pointed towards distinctly nuanced immune responses to different types of bacterial infection. Importantly, these immune signatures could be easily visualized to facilitate their interpretation, thereby allowing for rapid, accurate and transparent decision-making. This unique diagnostic capacity of Tsetlin Machines could help deliver early patient risk stratification and support informed treatment choices in advance of conventional microbiological culture results, thus guiding antibiotic stewardship and contributing to improved patient outcomes.

Availability and implementation: All underlying tools and the anonymized data underpinning this publication are available at https://github.com/anatoliy-gorbenko/biomarkers-visualization.