In recent years, omics technologies have largely driven the modernization of TCM ^[7]. Metabolomics is the apogee of the omics cascade that attempts to analyze a comprehensive set of metabolites in biological samples and explore changes in metabolic pathways related to genomic and proteomic perturbations ^[8]. TCM possesses several typical characteristics such as being multi-component, multi-target and multi-pathway, resulting in great difficulty when attempting to explore its pharmacological mechanisms ^[9]. Notably, metabolomics, especially untargeted metabolomics, can detect a global set of metabolites without bias in living organisms after TCM treatment, which provides the possibility of exploring the metabolic mechanisms of TCM in disease prevention and treatment ^[10]. Currently, two analytical platforms are mainly employed to acquire metabolomic data, including mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectroscopy ^[11]. These techniques have their advantages and disadvantages, as listed in Table 1. For example, the MS-based method has a higher sensitivity and more metabolites can be detected even using a minimal sample size. Moreover, the MS-based method is also a flexible technique, which can couple liquid or gas chromatography to achieve the selection and separation of different metabolites. However, it also has a number of disadvantages including low reproducibility, complex sample preparation, non-recyclable samples, relatively poor quantitative analysis and difficult metabolite identification. The NMR-based method possesses several strengths, such as high reproducibility, simple sample preparation, non-destructive, fast analysis, good quantitative analysis and straightforward identification, although this method cannot analyze non-protonated metabolites. In addition, compared with the MS-based method, NMR analysis needs a larger sample size and has a relatively low sensitivity. Figure 1 shows the typical ¹H NMR-based metabolomics profiling obtained from serum, liver and feces samples in healthy mice ^[12], and the detailed metabolite assignments are listed in Table 2, where a series of metabolites can be identified involving amino acid metabolism, energy metabolism, fatty acid metabolism, ketone body metabolism and others. NMR metabolomics profiling is tissue-specific due to different metabolite compositions. Therefore, different analytical sequences have been developed for NMR analysis. For example, a Carr–Purcell–Meiboom–Gill (CPMG) sequence is usually conducted for serum samples in order to minimize the line-broadening effect of blood macromolecules including proteins and lipids. However, for samples with a high water content such as urine, a standard single-pulse sequence (ZGPR) can be used to reduce the impact of water signals on metabolomics profiling. In addition, there is a greater likelihood of overlapping peaks from multiple metabolites with NMR analysis, resulting in difficult identification and quantification. One way to solve this problem is to perform NMR experiments under higher magnetic fields. Moreover, 2D J-resolved spectroscopy and spectral deconvolution have also been used to address the peak overlap of metabolites.

Figure 2 illustrates the flowchart of the NMR-based metabolomics method for elucidating the metabolic mechanisms of TCM for diabetes treatment. In brief, diabetic rodent models are treated with TCM after a period of time and then biological samples are collected for NMR-based metabolomics analysis, such as serum, plasma, urine, feces and tissue samples. Metabolomics data are subjected to multivariate and univariate analyses to identify important metabolites that are significantly altered after TCM treatment. Finally, metabolic pathway analysis is performed to elucidate potential therapeutic mechanisms of TCM for diabetes.