Machine learning discriminates bacterial, fungal, and viral infections using temporal blood analyte dynamics in bottlenose dolphins (Tursiops truncatus).

Temporal blood analyte trends analyzed via machine learning can reliably differentiate bacterial, fungal, and viral infections in dolphins.

This retrospective, longitudinal observational study evaluated whether machine learning models could differentiate bacterial, fungal, and viral infections in bottlenose dolphins (Tursiops truncatus) using temporal blood analyte dynamics. A total of 860 blood samples from 31 professional-care dolphins spanning 1995–2025 were analyzed across 33 confirmed disease episodes (11 bacterial, 10 fungal, 12 viral). Random forest models were trained on data from 5 defined temporal phases of infection, incorporating 34 key blood analytes. Model inputs included absolute analyte values, temporal slopes (rate of change over time), clinical ratios, and statistical aggregates. The overall model achieved 75.8% classification accuracy (25/33 episodes), with pathogen-specific performance of 80% for fungal, 75% for viral, and 72.7% for bacterial infections. Notably, temporal slope features—particularly the rate of change in eosinophil counts and total white blood cell counts—were selected in 96.8% of validation folds, underscoring the diagnostic superiority of dynamic versus static biomarker assessment. The authors conclude that longitudinal blood monitoring enables detection of subtle, individual-specific analyte shifts that outperform single time-point thresholds in pathogen discrimination. This study highlights the clinical value of personalized medicine approaches in marine mammal care, where individualized baseline data combined with machine learning may substantially improve early and accurate infectious disease diagnosis, guiding more targeted treatment decisions in a species where clinical signs can be subtle and diagnostics are challenging.

This summary was distilled by AI and may occasionally misinterpret data. Confirm critical details with the primary literature before clinical application.