A multimarker panel for diagnosis and prognosis prediction of fungal infections

Area of Science:

• Medical Diagnostics
• Biomarker Discovery
• Infectious Disease Research

Background:

• Fungal infections (FI) pose significant diagnostic challenges.
• Accurate diagnosis and prognostic stratification are crucial for effective patient management.
• Existing diagnostic methods may lack accessibility or comprehensive prognostic capability.

Purpose of the Study:

• To develop and validate a routine, indicator-based multimarker panel for fungal infection diagnosis.
• To predict the prognosis of patients with fungal infections.
• To create an accessible diagnostic and prognostic tool, particularly for resource-limited settings.

Main Methods:

• Retrospective cohort study of 134 patients, randomly split into training (n=75) and validation (n=59) sets.
• Screening of differential indicators using non-parametric tests and logistic regression, with LASSO regression for variable selection.
• Comparison of four machine learning algorithms (logistic regression, SVM, GNB, LightGBM) via 5-fold cross-validation, with the best model validated.
• Selection of six core indicators: hsCRP, BNP, PT%, D-Dimer, IL-17, and PCT.

Main Results:

• A logistic regression-based multimarker panel was developed, integrating inflammatory, coagulation, and cardiac biomarkers.
• The diagnostic performance in the validation set showed an AUC of 0.845 (sensitivity 80.00%, specificity 72.41%).
• The prognostic performance in the validation set yielded an AUC of 0.767 (sensitivity 72.41%, specificity 90.00%), outperforming single indicators.

Conclusions:

• The developed logistic regression-derived multimarker panel offers a reliable and accessible tool for fungal infection diagnosis.
• The panel provides effective prognostic stratification for patients with fungal infections.
• This approach is particularly valuable in resource-limited settings for improving patient care and outcomes.