Cocoa origin classifiability through LC-MS data: A statistical approach for large and long-term datasets.

Area of Science:

• Food Science
• Analytical Chemistry
• Chemometrics

Background:

• Food origin classification is crucial for quality assurance and flavor development.
• Liquid Chromatography-Mass Spectrometry (LC-MS) offers detailed chemical profiling of food.
• Large datasets present challenges due to experimental, instrumental, and batch variations.

Purpose of the Study:

• To address challenges in classifying food origin from complex LC-MS data.
• To evaluate Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) for cocoa origin classification.
• To develop an improved multivariate analysis approach for disparate datasets.

Main Methods:

• Utilized a dataset of 297 LC-MS profiles from cocoa sourced from 10 countries over four years.
• Applied Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA).
• Developed and tested a compound selection criterion based on Gaussian distribution of intensities.

Main Results:

• PCA showed limited separation of cocoa bean origins.
• Standard LDA exhibited non-linear dependencies on compound sets.
• The Gaussian-based compound selection criterion significantly enhanced origin clustering in LDA.
• The new approach improved multivariate analysis utility in complex, variable data.

Conclusions:

• Traditional multivariate methods have limitations for complex food origin classification.
• A novel compound selection strategy dramatically improves LDA performance for origin classification.
• This approach facilitates the study of marker compounds in large, heterogeneous food datasets.