Comparison of two exploratory data analysis methods for classification of Phyllanthus chemical fingerprint:...

Area of Science:

• * Analytical Chemistry
• * Cheminformatics
• * Pharmacognosy

Background:

• * Chemical fingerprinting is crucial for identifying and ensuring the quality of Traditional Chinese Medicines (TCM).
• * Phyllanthus species are widely used in TCM, necessitating robust analytical methods for their characterization.
• * Various classification methods exist, but a comprehensive comparison for TCM fingerprinting is lacking.

Purpose of the Study:

• * To holistically compare unsupervised and supervised pattern recognition technologies for analyzing Phyllanthus chemical fingerprints.
• * To evaluate the effectiveness of different methods in classifying Phyllanthus samples based on geographical origin and species.
• * To identify key chemical markers contributing to sample discrimination.

Main Methods:

• * High-performance liquid chromatography time-of-flight mass spectrometry (HPLC/TOFMS) was used to establish chemical fingerprints of 26 Phyllanthus samples.
• * Unsupervised methods (Principal Component Analysis - PCA) and supervised methods (Nearest Neighbors Algorithm - NN, Partial Least Squares Discriminant Analysis - PLS-DA, Artificial Neural Network - ANN) were employed.
• * Sixty-three compounds were identified and structurally assigned to build the fingerprints.

Main Results:

• * Artificial Neural Network (ANN) and Partial Least Squares Discriminant Analysis (PLS-DA) accurately classified Phyllanthus samples by geographical location and species.
• * Principal Component Analysis (PCA) successfully identified critical variables for discriminating between sample clusters.
• * Both unsupervised and supervised methods proved effective and reliable for TCM fingerprint analysis.

Conclusions:

• * Supervised methods like ANN and PLS-DA excel at sample classification in TCM fingerprinting.
• * Unsupervised methods like PCA are valuable for data mining and identifying key discriminatory variables.
• * Unsupervised and supervised pattern recognition technologies are complementary and enhance TCM chemical fingerprint analysis.