Rapid viral detection in milk using near infrared spectroscopy combined with Swiss knife partial least squares (SKPLS) 3-way multiblock

Human norovirus is the leading cause of foodborne illnesses, necessitating rapid, reliable, and cost-effective quantification methods. Near InfraRed Spectroscopy (NIRS) offers real-time analysis with minimal sample preparation. Two independent experiments, conducted 8 months apart in quadruplicate, produced 360 semi-skimmed milk measurement spiked with nine MS2 bacteriophage concentrations, a surrogate of human norovirus. Virus concentrations were modeled by 3-way partial least squares (PLS) regression correlating NIRS spectra with plaque-forming unit (PFU) measurements. A group-aware 5-fold cross-validation was applied to 72 sample groups using spectral pretreatments. The best performance was achieved with the extended multiplicative scatter correction (EMSC) and first derivative, yielding an R² of 0.77 and 0.80 for validation. The limit of quantification (LOQ) was estimated at 4.9 × 10³ PFU/mL, close to literature values for virus quantification in complex matrices. These results highlight the feasibility of NIRS as a rapid, non-destructive, and reagent-free approach for viral quantification in food systems.