Applications of front-face fluorescence spectroscopy and chemometrics to measure casein content in milk and detect protein leaks in dairy ultrafiltration permeates

Abstract

Quantitative analysis of casein content in cheese milk can give a better control over cheese yield and understand cheese quality. Traditional analytical methods for casein measurement rely on nitrogen-based quantification and involve time-consuming sample preparation steps. The current study applied front-face fluorescence spectroscopy (FFFS) combined with chemometrics to quantify casein and casein-to-crude-protein ratio (CN/CP) in milk intended for cheese manufacturing. FFFS spectra of acid-precipitated casein milk dispersion (pH = 4.6) were collected from 20 ultrafiltered (UF) model milk samples with different casein contents. A preliminary calibration model was developed with principal component regression (PCR) using reference casein contents and the FFFS spectra. The model was externally validated with 20 raw milk samples and a root mean square error (RMSE) of 0.15% was found between the predicted and reference casein contents. A relative prediction error (RPE) of 6.7% indicating usefulness for quality control purposes. To further refine the FFFS-based casein quantification method, 30 model milk samples prepared from UF and microfiltration (MF) permeates and retentates to generate different casein contents and CN/CP. The FFFS spectra were collected following the same procedure and used as predictors for casein and CN/CP quantifications. Calibration models were developed using partial least squares regression (PLSR) and elastic net regression (ENR) and the models were further optimized using 20 samples including raw, skim, and UF milk. The optimized PLSR and ENR models were again tested using 20 test samples including raw, skim, and UF milk and evaluated in terms of RMSE, residual prediction deviation (RPD), and RPE. The PLSR and ENR models reduced the RMSE for casein quantification to 0.13% with RPD ranged from 3.2 to 3.4, indicating practical model performances. For CN/CP quantification, PLSR models resulted in useful predictions with an RMSE of 0.024, an RPD of 1.5, and an RPE of 3.0%. The FFFS-based casein quantification method provides a rapid casein measurement in fluid milk and can be implemented in the cheese industry for routine measurements. In a different study, FFFS was utilized to predict the protein leaks in permeate during membrane processing of skim milk and whey. Protein leak occurs when proteinous matters pass through the UF membrane into the permeate stream leading to financial losses and product quality defects. FFFS as a sensitive and specific instrument was applied to characterize protein leak occurrences in UF permeate, develop chemometrics models to quantify true protein (TP) content in permeate streams, and classify sources of protein leak in the feed material. Measurements of crude protein (CP), non-protein nitrogen, TP, tryptone-equivalent peptide, α-lactalbumin (α-LA), and β-lactoglobulin (β-LG) were performed on 33 lots of commercial whey permeate and 29 lots of commercial milk permeate. Protein leaks were attributed to high TP, high-peptide, and presence of α-LA or β-LG. Tryptophan was identified as the fluorophore of interest for protein leak detection based on the excitation-emission matrix analysis of representative permeate with high and low TP contents. Quantitative models based on PLSR were developed using tryptophan excitation spectra and true protein content in the permeate. The model yielded a RMSE of 0.22% (dry-basis) and RPD of 2.8 based on external validations, showing a useful model for quality control purposes. Moreover, classification models based on partial least squares discriminant analysis were developed to detect high TP level, high peptide level, and presence of α‐LA or β-LG with 83.3%, 84.8%, and 98.5% cross-validated accuracy, respectively. This method showed that FFFS and chemometrics can rapidly detect protein leak and identify the source of protein leak in UF permeate, which can reduce financial loss from protein leak and maintain high-quality permeate production.