Detecting Misrepresentation, Adulteration in Cheese

Today, routine checks for incoming materials are often done by near infrared spectroscopy (NIR) or fourier transform infrared spectrometry, or FTIR. These are useful techniques as they assess profiles instead of single parameters. NIR technology focuses mainly on three major chemical entities: C-H, N-H, O-H, and C-O-H, representing sugar, water, protein, and fat. This is insufficient to identify all adulterants or changes unrelated to any of these structures.

Novel Methods of Detection

Nowadays, cases of fraud tend to be substantially more sophisticated, and scientists are behind some of them. In many instances fraudsters take advantage of the limitation of detection methods or by the fact that many compounds are not normally associated with foods and therefore are not looked for. Melamine is a very good example: In order to perform this fraud and keep it covered, it is important to understand that one of the quality parameters for milk, protein content, is not measured directly. Instead, it is assessed by methods (Kjeldahl and Dumas), which determine nitrogen content. Such methods do not only determine nitrogen in the protein structure, but also nitrogen in other compounds present in the sample. Therefore, more nitrogen does not necessarily translate into more protein and thus higher quality.

Since many fraudsters target methods of analysis, it is critical to develop new strategies to counteract. One good option is the use of novel technologies that allow the simultaneous assessment of a wide range of different variables, which in their entirety, is difficult to fool.

Figure 1

Image credit: Mérieux NutriSciences

Among the novel technologies worth mentioning is the combination of high-resolution mass spectrometry with sophisticated multivariate statistical analysis. The data generated by the mass spectrometric analysis are processed by software that generates a three dimensional model (see Figure 1), which looks like a sphere. A key precondition for developing these models is to have a certain number of reference samples known to be authentic. The non-targeted approach that Mérieux NutriSciences has developed to verify the authenticity of Parmigiano Reggiano was built on one reference sample provided by the Parmigiano Reggiano Consortium.

After the model has been built, unknown samples are analyzed and compared with the multi-variable model. The model will distinguish samples that are compliant (authentic Parmigiano Reggiano) from those which are not. If not, there is a high probability that a sample is adulterated or mislabeled.

In case of the Parmigiano Reggiano model, the high-resolution non-targeted mass spectrometry in conjunction with statistics already provided a good prediction rate. This could be further improved by assessing additional targeted variables, e.g. compound only present in silage feed. This approach yielded a 100 percent prediction rate when tested on blind samples sent by the Parmigiano Reggiano Consortium.

The principle of this non-targeted approach can easily be transferred to other premium products as long as reference materials are available. In the case of extra virgin olive oil, not only the country and region could be predicted, but also the year of the olive harvest.

The work demonstrates that older and costly approaches do not always lead to better or more accurate results. The novel non-targeted approach, based on numerous examples, can result in ideal prediction of adulterated or mislabeled sample.

Popping, is chief scientific officer at Mérieux NutriSciences Corp. Reach him at [email protected]. And de Dominicis is head of research and development at Mérieux NutriSciences in Italy. Reach him at [email protected].