One-class modeling for verification of botanical identity: a review

Area of Science:

• Chemometrics
• Botanical Identification
• Multivariate Analysis

Background:

• Botanical product authentication is critical for quality control and consumer safety.
• Traditional identification methods can be subjective or lack comprehensive analytical power.
• Multivariate statistical methods offer robust approaches for complex sample analysis.

Purpose of the Study:

• To review multivariate, one-class modeling based on Principal Component Analysis (PCA) for botanical identification.
• To highlight the flexibility and utility of one-class modeling compared to other methods.
• To demonstrate the application of one-class modeling using real-world botanical examples.

Main Methods:

• Focus on supervised multivariate statistical methods, specifically one-class modeling.
• Utilize Principal Component Analysis (PCA) to build a reference model from authentic samples.
• Employ the Q statistic as a combined metric to assess test sample similarity to the model.
• Discuss factors influencing identification: number of variables, classes, and analysis type (quantitative/qualitative).

Main Results:

• One-class modeling provides a flexible and effective approach for botanical identification.
• Multivariate analysis offers broader coverage of sample characteristics compared to univariate methods.
• The Q statistic effectively classifies samples as authentic or non-authentic based on model limits.
• Demonstrated successful application in identifying American ginseng, Echinacea purpurea, Black Cohosh, and Maca.

Conclusions:

• Multivariate one-class modeling using PCA is a powerful tool for botanical authentication.
• The method's flexibility allows for adaptation to various botanical identification challenges.
• This approach enhances the reliability and accuracy of identifying plant species and detecting adulteration.