Centrifugation-induced somatic cell count reduction in milk as a form of adulteration detected by 1H nuclear magnetic resonance spectroscopy and multivariate modeling.

Milk adulteration continues to pose significant challenges in the 21st century, primarily due to the inventiveness of fraudsters, as evidenced by a newly reported method of falsifying SCC through centrifugation. Despite its potential relevance, this form of SCC adulteration remains largely unexplored. This study evaluated the potential of proton nuclear magnetic resonance (1H NMR) spectroscopy combined with chemometric techniques to detect the artificial reduction of SCC in bovine milk caused by centrifugation. Raw bulk milk samples (n = 56) originating from 4 Czech dairy farms, including 2 Holstein and 2 Czech Fleckvieh herds, were analyzed before and after centrifugation, which resulted in an average relative SCC reduction of 52.2%. Milk quality parameters, including SCC, electrical conductivity, pH, freezing point, and lactose, were assessed and used as complementary variables alongside 1H NMR spectral data. Among the classification models tested, the generalized linear model with lasso regularization (GLM-Lasso) achieved a classification accuracy of 91.96% and yielded the highest area under the curve (AUC) values when using an SCC threshold of 200 × 103 cells/mL and incorporating both spectral and physicochemical variables. Other models, including support vector machines, partial least squares-discriminant analysis, and the generalized linear model with ridge regularization, achieved accuracies ranging from 77.68% to 91.96% but yielded lower AUC values. Originally elevated SCC was associated with multiple NMR spectral signals, particularly those assigned to choline-containing compounds, amino acids, and organic acids. The centrifugation step did not introduce unique measurable changes in 1H NMR spectra, as evidenced by the limited classification accuracy (55%) between groups of spectra with and without the centrifugation step. This preservation of the SCC-associated metabolic profile enabled the identification of discrepancies between declared SCC values and the corresponding metabolic profile of milk. These findings demonstrate the potential of 1H NMR spectroscopy combined with GLM-Lasso for detecting centrifugation-induced SCC reduction in bovine milk.