A model for predicting bacteremia species based on host immune response
Peter Simons1, Virginie Bondu1, Laura Shevy2
1Department of Pathology, University of New Mexico Health Sciences Center, Albuquerque, NM, United States.
Frontiers in Cellular and Infection Microbiology
|March 5, 2025
Summary
This study developed a machine-learning model combining immune cell activation and blood count data to rapidly identify bacterial pathogens in patients with bacteremia, enabling faster treatment decisions.
Area of Science:
- Immunology
- Computational Biology
- Clinical Microbiology
Background:
- Bacteremia diagnosis and antibiotic selection are challenging for clinicians.
- Rapid identification of bacterial species is crucial for timely treatment.
Purpose of the Study:
- To develop a novel approach for rapid bacterial pathogen identification in bacteremia.
- To combine host immune response data with machine learning for clinical decision support.
Main Methods:
- Quantitative measurement of Rho family GTPase activity (Rac1•GTP) using a novel 'G-Trap assay'.
- Analysis of leukocyte populations from routine complete blood counts with differential.
- Development of a machine-learning model using partial least squares discriminant analysis (PLS-DA).
Main Results:
- Increased Rac1•GTP levels were observed in 18 of 28 bacteremia patients.
- Some bacteremia patients showed immunosuppression patterns with normal Rac1 GTPase activity.
- PLS-DA model effectively differentiated specific pathogen groups with validated predictive utility.
Conclusions:
- Combining routine blood assays with machine learning offers a potential clinical decision-support tool.
- This approach could enable prompt, accurate, same-day treatment for bacteremia.
- Rapid pathogen identification via immune response profiling is feasible.
