Authentication and geographical origin discrimination of Citri Reticulatae Pericarpium via ICP-MS-based multielement profiling and interpretable machine learning

To accurately and reliably discriminate Guang Chenpi (GCP) from Chenpi (CP), two commercial types of Citri Reticulatae Pericarpium (CRP), we report an authentication strategy based on inductively coupled plasma mass spectrometry (ICP-MS) multielement profiling combined with interpretable machine learning (ML). Twenty-four mineral elements were quantified in 96 CRP samples from four producing regions in China via ICP-MS. Principal component analysis and orthogonal partial least-squares discriminant analysis revealed regional trends but could not completely separate GCP from CP. Subsequently, nine supervised ML models were designed. Among them, five models demonstrated particularly strong classification performance. Integrating model performance and feature-importance rankings, five mineral elements (V, Zn, Na, K, and Cu) were identified as elemental quality markers for origin authentication. Models developed using these markers retained excellent predictive power, with accuracies of > 0.963 in both internal and external validation sets. These results indicate that ICP-MS-based elemental profiling combined with interpretable ML provides an effective and practical approach for CRP authenticity evaluation, geographical traceability and quality control.