Integrated machine learning-based approach to evaluate authenticity in various food matrices via MALDI-TOF-MS technology

Significance of the Topic

The integrity of food products is critical to consumer trust, safety, and economic value. Authenticity assessment of items such as olive oil and dairy has gained momentum to combat fraud and ensure quality. Integrating machine learning with MALDI-TOF-MS provides a rapid, reliable screening tool for traceability and verification in complex food matrices.

Objectives and Study Overview

This study introduces a holistic workflow combining MALDI-TOF-MS lipid profiling with deep learning to determine the origin and detect adulteration in olive oil and milk. Case studies include extra virgin olive oil versus refined and various milk species, with sensitivity down to 1% adulteration.

Materials and Methods

Sample preparation involved chloroform–methanol extraction and application of α-cyano-4-hydroxycinnamic acid as the MALDI matrix. A double-layer spotting technique on a stainless-steel target plate was employed. Data acquisition was performed on a Bruker microflex LRF MALDI-TOF instrument in reflector mode. The raw spectra were processed into total ion chromatograms, converted to greyscale images, and encoded into fingerprint bits for model input.

Instrumental Setup

• MALDI-TOF-MS system: Bruker microflex LRF, reflector mode
• MALDI matrix: α-cyano-4-hydroxycinnamic acid
• Sample spotting: double-layer on ground steel target

Key Results and Discussion

The convolutional neural network achieved high classification accuracy for oils (extra virgin, refined, sunflower) and milks (cow, goat, sheep), reliably identifying sample origin. Adulteration levels as low as 1% were detected. Probability outputs can be encoded into a digital mark for direct consumer or manufacturer feedback and enhanced traceability.

Practical Benefits and Applications

• Rapid screening with minimal sample preparation
• Low per-sample cost and high throughput
• Universal approach adaptable to diverse food matrices
• Digital mark (e.g., QR code) integration for transparency

Future Trends and Opportunities

Expansion to additional food products, real-time on-site analysis, and integration with production-line monitoring are envisaged. Advances in machine learning algorithms and miniaturized MALDI-TOF platforms will further enhance speed and accuracy, supporting broader implementation in quality assurance and regulatory frameworks.

Conclusion

The developed MALDI-TOF-MS workflow combined with deep learning offers a fast, cost-effective, and reliable method for food authenticity assessment, demonstrating high sensitivity for origin determination and adulteration detection. This platform supports enhanced traceability and consumer confidence.

References

1. Food Chemistry, DOI: 10.1016/j.foodchem.2021.131057
2. Food Chemistry, DOI: 10.1016/j.foodchem.2012.02.154