Machine Learning–Driven Near‐ and Mid‐Infrared Chemometrics for Rapid, Cost‐Effective COVID‐19 Screening in Dried Plasma

ABSTRACT The COVID‐19 pandemic has highlighted the urgent need for rapid, accurate, and cost‐effective diagnostic alternatives to conventional methods such as RT‐qPCR and immunoassays. In this study, we explored the potential of vibrational spectroscopy, specifically near‐infrared (NIR) and mid‐infrared (MID) spectroscopy, for detecting SARS‐CoV‐2 infection in dried plasma samples. Spectral data were obtained from 83 patients (45 COVID‐19 positive and 38 negative) and analyzed using partial least squares discriminant analysis (PLS‐DA), with and without variable selection by the ordered predictors selection for discriminant analysis (OPSDA) method. While initial models using full spectra showed moderate classification accuracy, the application of OPSDA significantly enhanced model performance. For the NIR dataset, OPSDA‐based models achieved 100% sensitivity and specificity in both training and test sets ( n = 25). For the MID dataset, the test set ( n = 25) sensitivity reached 86%, with 100% specificity. These results demonstrate that NIR and MID spectroscopy, when combined with advanced chemometric approaches, can provide reliable, rapid, and low‐cost screening for COVID‐19. This platform holds promise for broader applications in clinical diagnostics beyond the current pandemic.