Cocoa bean fingerprinting via correlation networks

Area of Science:

• Food science
• Analytical chemistry
• Network science

Background:

• Cocoa bean fermentation lacks standardization, complicating food authenticity and quality assessments.
• Existing methods struggle with the chemical complexity of cocoa products.

Purpose of the Study:

• To demonstrate how network science can enhance food fingerprinting and authenticity research for cocoa.
• To develop a data-driven approach for classifying cocoa based on its chemical profile.

Main Methods:

• Utilized liquid chromatography-mass spectrometry (LC-MS) to profile 140 cocoa samples.
• Constructed correlation networks based on LC-MS data to analyze sample relationships.
• Applied network visualization, statistics, and machine learning for data interpretation.

Main Results:

• Correlation network topology successfully distinguished cocoa samples by fermentation/processing stage and country of origin.
• Increasing correlation thresholds revealed hierarchical clustering, first by processing stage, then by country.
• The network-based approach provided both qualitative and quantitative classification evidence.

Conclusions:

• Network science offers a powerful tool for analyzing complex mass spectrometry data in food authenticity studies.
• This approach can effectively fingerprint cocoa, differentiating processing stages and origins.
• The methodology shows potential for broad application in food analysis beyond cocoa.