Leveraging Multi-Level Biomarkers Using Machine Learning: Identifying Physiological and Skin Microbial Dynamics in...

Area of Science:

• Amphibian ecology and disease dynamics
• Microbiology and immunology
• Machine learning applications in wildlife health

Background:

• Global amphibian declines are driven by stressors, including chytridiomycosis caused by Batrachochytrium dendrobatidis (Bd).
• South Korean amphibian species like Pelophylax nigromaculatus exhibit resistance to Bd, complicating the assessment of its precise impacts.
• Understanding Bd infection dynamics in resistant species is crucial for amphibian conservation efforts.

Purpose of the Study:

• To investigate the dynamics of Batrachochytrium dendrobatidis (Bd) infection in Pelophylax nigromaculatus.
• To integrate physiological, microbiological, and morphological data with machine learning for infection analysis.
• To identify reliable biomarkers for Bd infection status in a Bd-resistant amphibian population.

Main Methods:

• Collected data on Bd prevalence, body size, weight, corticosterone (CORT) levels, innate immune function (bacterial killing assays), and skin microbiome composition.
• Applied machine learning methodologies, including Light Gradient Boosting Machine and generative adversarial networks for data augmentation.
• Analyzed physiological and microbiome data to differentiate between infected and non-infected individuals.

Main Results:

• Significant physiological differences were observed between infected and non-infected P. nigromaculatus, notably elevated CORT levels and altered bacterial killing capacity.
• Skin microbiome analysis revealed subtle variations but no significant differences in alpha or beta diversity between infected and non-infected groups.
• Machine learning models, particularly Light Gradient Boosting Machine with data augmentation, achieved high predictive performance for classifying infection status.

Conclusions:

• Corticosterone levels and bacterial killing ability are effective predictors of Bd infection status in Pelophylax nigromaculatus.
• Machine learning can effectively integrate multi-level biomarkers to assess amphibian health, even with variations in disease resistance and infection loads.
• This integrated approach is vital for understanding and mitigating the impacts of multiple threats on amphibian populations worldwide.