Sixty peony root training samples of the same age were collected from various regions in Korea and China, and their genetic diversity was investigated for 23 chloroplast intergenic space regions. All samples were genetically indistinguishable, indicating that the DNA-based techniques employed were not appropriate for determining the samples' regions of origin. In contrast, (1)H-nuclear magnetic resonance ((1)H-NMR) spectroscopy-based metabolomics coupled with multivariate statistical analysis revealed a clear difference between the metabolic profiles of the Korean and Chinese samples. Orthogonal projections on the latent structure-discrimination analysis allowed the identification of potential metabolite markers, including γ-aminobutyric acid, arginine, alanine, paeoniflorin, and albiflorin, that could be useful for classifying the samples' regions of origin. The validity of the discrimination model was tested using the response permutation test and blind prediction test for internal and external validations, respectively. Metabolomic data of 21 blended samples consisting of Korean and Chinese samples mixed at various proportions were also acquired by (1)H-NMR analysis. After data preprocessing which was designed to eliminate uncontrolled deviations in the spectral data between the testing and training sets, a new statistical procedure for estimating the mixing proportions of blended samples was established using the constrained least squares method for the first time. The predictive procedure exhibited relatively good predictability (adjusted R (2) = 0.7669), and thus has the potential to be used in the quality control of peony root by providing correct indications for a sample's geographical origins.