Hepatitis C Virus positivity prediction from serum samples using NIRS and L1-penalized classification

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Hepatology

Background:

• Hepatitis C virus (HCV) infection leads to diverse pathological outcomes, including fibrosis, cirrhosis, and hepatocellular carcinoma (HCC).
• Predicting the evolutionary course of HCV infection remains challenging due to complex host-pathogen interactions.
• Current diagnostic methods for HCV positivity can be further enhanced with advanced analytical techniques.

Purpose of the Study:

• To develop a predictive model for discriminating HCV positivity in patient serum samples.
• To explore the combined potential of Near-infrared spectroscopy (NIRS) and machine learning for HCV detection.
• To establish a foundation for predicting HCV infection progression.

Main Methods:

• Utilized 126 serum samples from 38 HCV patients at various disease stages.
• Acquired Near-infrared spectroscopy (NIRS) spectra using a Perkin Elmer FT-NIRS Spectrum 100 device.
• Employed an L1-penalized logistic regression model with 5-fold cross-validation to classify HCV presence based on spectral data.

Main Results:

• The L1-penalized logistic regression model identified 167 significant wavelengths.
• Achieved an accuracy of 0.78 on an independent test set for HCV positivity detection.
• Demonstrated the model's ability to classify serum samples as positive or negative for HCV presence.

Conclusions:

• Presented a straightforward and promising approach combining NIRS and machine learning for HCV detection in serum samples.
• The findings encourage further development for predicting HCV progression and other clinical applications.
• Highlighted the clinical relevance of using machine learning and NIRS for analyzing viral presence in biological samples.