Development of Non-Targeted Mass Spectrometry Method for Distinguishing Spelt and Wheat.

Area of Science:

• Food Science
• Analytical Chemistry
• Computational Biology

Background:

• Food fraud is a pervasive issue requiring novel detection strategies.
• Authenticity testing is crucial for consumer trust and regulatory compliance.
• Distinguishing between similar grains like spelt and wheat can be challenging.

Purpose of the Study:

• To develop and validate a non-targeted method (NTM) for differentiating spelt and wheat.
• To enhance food fraud detection and authenticity testing capabilities.
• To assess the reliability of the NTM across diverse cultivars and processed goods.

Main Methods:

• Utilized liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS) to generate spectral fingerprints.
• Developed convolutional neural network (CNN) models employing a nested cross-validation (NCV) approach.
• Trained models on calibration sets of wheat and spelt cultivars, then validated with external datasets including mixtures and processed goods.

Main Results:

• CNN models successfully learned spectral patterns to discriminate between spelt and wheat samples.
• The NTM demonstrated reliability on external validation sets, including artificially mixed spectra and processed food items.
• A new metric, the D score, was introduced for quantitative evaluation of classification decisions.

Conclusions:

• The developed NTM, integrating NCV and CNNs with spectral data, offers a reliable solution for food fraud detection.
• This method is robust enough for application across a broader range of cultivars and their mixtures.
• The findings support the use of advanced analytical and computational techniques for ensuring food authenticity.