Lecture
Authentication of food using analytical fingerprints and chemometrics
- at -
- ICM Saal 4b
- Type: Lecture
Lecture description
Food fraud is a widespread problem involving the deliberate adulteration of food products for financial gain. It can endanger consumer safety, damage trust and result in substantial financial losses. In order to detect food adulteration, the use of powerful analytical methods is essential. These methods usually involve obtaining an analytical fingerprint that can be compared with those of authentic samples. This requires a compromise between the quick and easy application, which is possible for example using handheld spectrometers, and the most thorough capture of the analytical fingerprint of the sample. The latter can be improved by combining multiple approaches.
In order to comprehensively exploit the complex data generated by analytical methods, multivariate chemometric methods are applied. These methods are divided into unsupervised approaches, which are applied without further information, and supervised
approaches, by which classification models are trained using samples with known class memberships. Some of the latter, machine learning approaches, can be applied as black boxes meaning that only the class assignment is reported, while the background that led to this decision remains unknown. However, it is important to clarify how the corresponding supervised models work, i.e., to analyze which variables in the data set are relevant and how they interact. We have developed and implemented multiple methods for selecting important variables and analyzing relationships between them in random forest models [1,2].
This presentation introduces a variety of methods for authenticating food. Some aim to facilitate easy application using handheld devices, while others exploit comprehensive analytical fingerprints that can be characterized through variable selection and variable relation analysis.
References:
[1] S. Seifert, S. Gundlach, S. Szymczak, Bioinformatics 2019, 35, 3663-3671.
[2] L. Voges, L. Jarren, S. Seifert, Bioinformatics, 2023, 39, btad471.
In order to comprehensively exploit the complex data generated by analytical methods, multivariate chemometric methods are applied. These methods are divided into unsupervised approaches, which are applied without further information, and supervised
approaches, by which classification models are trained using samples with known class memberships. Some of the latter, machine learning approaches, can be applied as black boxes meaning that only the class assignment is reported, while the background that led to this decision remains unknown. However, it is important to clarify how the corresponding supervised models work, i.e., to analyze which variables in the data set are relevant and how they interact. We have developed and implemented multiple methods for selecting important variables and analyzing relationships between them in random forest models [1,2].
This presentation introduces a variety of methods for authenticating food. Some aim to facilitate easy application using handheld devices, while others exploit comprehensive analytical fingerprints that can be characterized through variable selection and variable relation analysis.
References:
[1] S. Seifert, S. Gundlach, S. Szymczak, Bioinformatics 2019, 35, 3663-3671.
[2] L. Voges, L. Jarren, S. Seifert, Bioinformatics, 2023, 39, btad471.
