Cocoa quality: Chemical relationship of cocoa beans and liquors in origin identitation

• Published in: Food research international Vol. 172; p. 113199
• Main Authors: Bagnulo, Eloisa, Scavarda, Camilla, Bortolini, Cristian, Cordero, Chiara, Bicchi, Carlo, Liberto, Erica
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.10.2023
• Subjects: Cocoa beans and liquors; Fingerprinting; Machine learning; Origin identitation; Profiling; Quality; Origin identitation; Profiling; Fingerprinting; Quality; Machine learning; Cocoa beans and liquors
• ISSN: 0963-9969; 1873-7145
• DOI: 10.1016/j.foodres.2023.113199

[Display omitted] •Cocoa volatilome of beans and liquors is a tool for origin identitation and authentication.•Identitation at the molecular level is an objective way to qualify flavour in long time studies.•Machine learning affords to define origin classification model for the chemical-sensory identitation.•Fingerprinting and profiling approaches provided comparable classification performances.•Targeted approach is mandatory when a quality certification of origin is requested. In this study, HS-SPME-GC–MS was applied in combination with machine learning tools to the identitation of a set of cocoa samples of different origins. Untargeted fingerprinting and profiling approaches were tested for their informative, discriminative and classification ability provided by the volatilome of the raw beans and liquors inbound at the factory in search of robust tools exploitable for long-time studies. The ability to distinguish the country of origin on both beans and liquors is not so obvious due to processing steps accompanying the transformation of the beans, but this capacity is of particular interest to the chocolate industry as both beans and liquors can enter indifferently into the processing of chocolate. Both fingerprinting (untargeted) and profiling (targeted) strategies enable to decipher of the information contained in the complex dataset and the cross-validation of the results, affording to discriminate between the origins with effective classification models.