PLS-DA vs sparse PLS-DA in food traceability. A case study: Authentication of avocado samples.

Area of Science:

• Food science and analytical chemistry, focusing on chemometrics and spectroscopic analysis.

Background:

• Accurate authentication of food products, like avocados, is crucial for quality control and preventing fraud.
• Distinguishing between different geographical origins and cultivars of avocados presents a significant analytical challenge.

Purpose of the Study:

• To develop and validate robust classification models for authenticating avocado samples based on their geographical origin and cultivar.
• To evaluate the effectiveness of conventional and sparse partial least squares-discriminant analysis (PLS-DA and sPLS-DA) in multiclass food authentication problems.

Main Methods:

• Lipid chromatographic fingerprints of avocado samples were acquired using gas chromatography-flame ionization detection (GC-FID).
• Multivariate classification models, including PLS-DA and sPLS-DA, were built using the GC-FID data.
• A classification model concatenating strategy was employed to enhance the resolution of multiclass authentication problems.

Main Results:

• Both PLS-DA and sPLS-DA demonstrated successful authentication of avocado samples across three geographical origins and six cultivars.
• The classification models achieved high performance metrics, with accuracy values around 0.95.
• The concatenating strategy proved effective in resolving complex multiclass authentication challenges.

Conclusions:

• GC-FID combined with PLS-DA and sPLS-DA offers a reliable approach for avocado authentication.
• The developed methods can accurately differentiate avocado samples by geographical origin and cultivar.
• This study highlights the potential of chemometric techniques in ensuring food authenticity and traceability.