Info

a Determined by sensory panel.

b Predicted from SPME-MS-MVA data using PLS prediction models. c Error = predicted — actual. d Average error = (X| error|)/n, where n = 20.

a Determined by sensory panel.

b Predicted from SPME-MS-MVA data using PLS prediction models. c Error = predicted — actual. d Average error = (X| error|)/n, where n = 20.

Actual Shelf Life

Figure 13 Plot of predicted shelf life (based on PLS modeling) versus actual shelf life (based on sensory testing) for 64 samples used to prepare a PLS model for reduced-fat milk. Analyses performed by SPME-MS-MVA.

Actual Shelf Life

Figure 13 Plot of predicted shelf life (based on PLS modeling) versus actual shelf life (based on sensory testing) for 64 samples used to prepare a PLS model for reduced-fat milk. Analyses performed by SPME-MS-MVA.

Figure 13 shows a plot of predicted shelf life versus actual shelf life (based on sensory testing) for the 64 samples used to prepare the PLS model for reduced-fat milk. Table 3 shows PLS statistics for the PLS shelf-life prediction model and the PLS predictions of Model Validation Subset samples.

Preliminary results using SPME-MS-MVA as an electronic-nose system appear to give more accurate predictions of milk shelf life than most methods currently used to estimate milk shelf life. (See Table 4.) SPME-MS-MVA is also faster and easier to implement than other shelf-life prediction methods. Furthermore, SPME-MS-MVA has been shown to be useful for identifying samples with nonmicrobiological induced off-flavors and for determining the cause of off-flavors even when nonmicrobiological agents are involved.

Over a 7-month period, SPME-MS-MVA has been shown to be an accurate technique for predicting the shelf life of reduced-fat milk. Despite the fact that during the testing period significant changes occurred with the mass spectrometer (replacement of the turbomolecular pump and replacement of the electron multiplier) and the fact that several different Carboxen/PDMS fibers were used, internal standard normalization with chlorobenzene allowed accurate prediction over the 7-month period. Long-term stability, a problem with many e-nose instruments based on solid-state sensors, does not appear to be a significant problem with MS-based e-nose instruments.

Using Carboxen/PDMS SPME fibers to extract volatiles offers impressive

Table 3 PLS Error Analysis for Calibration Model and Model Validation Subset for Reduced-Fat Milk (shelf life in days as dependent variable)

Calibration Model validation model subset

PRESSa 53.4204 25.5115

SECb (days) 1.0441

SEPc (days) 1.1294

R2 0.9882 0.9801

Factors 15 7

Slope 0.9766 0.9492

Intercept (days) 0.2014 0.0750

n 64 20

a PRESS = Prediction Residual Error Sum of Squares. b SEC = Standard Error of Calibration = (PRESS/n)1/2. c SEP = Standard Error of Prediction = [PRESS/(n - k)]1 where k = number of factors.

advantages over static headspace (SH) and dynamic headspace (DH) sampling techniques. It does not require expensive ancillary instrumentation and is far more efficient than either SH or DH at extracting volatile fatty acids (VFAs) from milk. VFAs, important contributors to malodors and off-flavors in milk, are generated as metabolites by the growth of lipolytic psychrotrophic bacteria. Malodorous VFAs are too polar to detect at low levels using SH and DH as sample preparation/extraction tools.

Table 4 Correlation Coefficients for Various Shelf-Life Prediction Tests

Correlation coefficient

Shelf-life prediction test (predicted vs. actual)

Moseley Keeping-Quality Test 0.7-0.77

Catalase 0.77

DEFTa 0.72

VTSLb 0.89

SPME-MS-MVA 0.98

a Direct Epifluorescent Filter Technique. b Virginia Tech Shelf-Life Program.

Source: Dr. Russel Bishop, University of Wisconsin, Madison, Wisconsin.

SPME-MS-MVA has strong potential applications in the dairy industry for shelf-life prediction. Testing over a longer period of time and sampling from different production facilities should be conducted to confirm the accuracy of this new test as a predictor of shelf life. With a SPME autoinjector and minor test modifications, it would be possible to analyze one sample every five to seven minutes. The only labor required by the QC technician would be to pipette 3 mL of milk into a GC vial.

A simple, rapid, and sensitive sample preparation technique such as SPME and applying GC/MS in a nontraditional way are the basis of this new shelf-life prediction test for milk. This approach could be extended to many other types of important quality control applications in the food industry as well as in other industries.

0 0

Post a comment