Mass spectrometry data on specialized metabolome of medicinal plants used in East Asian traditional medicine

Traditional East Asian medicine not only serves as a potential source of drug discovery, but also plays an important role in the healthcare systems of Korea, China, and Japan. Tandem mass spectrometry (MS/MS)-based untargeted metabolomics is a key methodology for high-throughput analysis of the complex chemical compositions of medicinal plants used in traditional East Asian medicine. This Data Descriptor documents the deposition to a public repository of a re-analyzable raw LC-MS/MS dataset of 337 medicinal plants listed in the Korean Pharmacopeia, in addition to a reference spectral library of 223 phytochemicals isolated from medicinal plants. Enhanced by recently developed repository-level data analysis pipelines, this information can serve as a reference dataset for MS/MS-based untargeted metabolomic analysis of plant specialized metabolites. Measurement(s) Metabolomics Technology Type(s) high-performance liquid chromatography-electrospray ionisation tandem mass spectrometry Sample Characteristic - Organism Embryophyta Measurement(s) Metabolomics Technology Type(s) high-performance liquid chromatography-electrospray ionisation tandem mass spectrometry Sample Characteristic - Organism Embryophyta


Background & Summary
Most cultures worldwide use plants to treat diseases. The integration of experimental knowledge on medicinal plant usage with theories or beliefs about health and illness is termed traditional medicine. Traditional East Asian traditional medicine is known to have originated approximately 3,000 years ago in China. It was introduced to Korea and Japan from China in the 6th century, with Buddhism and Chinese culture 1 . Since then, it has been widely used following a long history. The detailed practices are not exactly same in three countries due to several reasons, for example, the usage of different species in the same genus due to the different climate; however, they have strongly influenced each other. Traditional East Asian medicine still plays an essential role in public health care in East Asian countries; currently, standardized herbal formulae are manufactured by pharmaceutical companies and used as parts of the modern medical systems in Korea, China, and Japan. Medicinal plants used in traditional medicine are one of the most important sources of drug discovery, where artemisinin, an antimalarial agent from Artemisia annua, is the most representative case.
High-throughput analysis of samples with complex chemical compositions plays a key role in the investigation and modernization of traditional East Asian medicine. Tandem mass spectrometry (MS/MS), especially in combination with liquid chromatography (LC), is the analytical method most commonly used to analyze medicinal plants 2,3 . MS/MS-based untargeted metabolomics has been used to assess the quality of medicinal plants and related dietary and pharmaceutical products, and has also been utilized in structure-based bioactive compound discovery [4][5][6][7] . Despite the increased use of this technique, only a few reliable and controlled datasets of MS/MS data for medicinal plants have been deposited in public repositories. With the expansion of untargeted metabolomics to multiple fields, the importance of publicly available data is increasing. The successive launches of MASST and ReDU symbolize the increasing need for public datasets in MS/MS-based untargeted metabolomics research 8,9 . MASST enables the search for a single spectrum by comparison with all publicly available raw data, whereas ReDU enables the reuse of deposited datasets for repository-level analyses or co-analysis with the user's own experimental data.
This Data Descriptor documents publicly available and re-analyzable raw LC-MS/MS dataset of 337 medicinal plants on the MassIVE raw data repository, which is linked with the Global Natural Product Social Molecular Networking platform (GNPS, https://gnps.ucsd.edu) 10 . This dataset is referred to as the KP337 dataset in this Data Descriptor, as most of the medicinal plants in the dataset are listed in the Korean Pharmacopeia (KP). The data do not cover the entire set of medicinal plants enlisted in KP, but cover most of the commonly used plants. The taxonomic coverage of the plants and plant parts used in Traditional East Asian medicine is summarized in Fig. 1. The KP337 dataset consists of raw LC-MS/MS data acquired in both positive and negative ion mode, and metadata formatted for compatibility with ReDU 9 . Thus, this dataset enables data re-usage, such as comparative analysis or propagation of spectral annotation based on spectral similarity 11 . Recently, a part of this dataset, specifically relating to various flavonoid C-glycosides, was utilized to establish the GNPS nearest neighbor suspect spectral library 12 . This case demonstrates the applicability of public datasets for propagation of spectral annotations. The KP337 dataset can also be applied to a MASST search of MS/MS spectra, which suggests the possible occurrence of queried molecules in medicinal plants. In natural product discovery projects, known compounds are often ignored and not reported. However, novel occurrences of known chemicals can provide insights into the medicinal or biological properties of medicinal plants, where the present dataset can contribute to such findings via MASST. Additionally, the occurrence data of known or unknown chemicals can enhance reference data-driven analysis, which was suggested as an alternative workflow for MS/MS-based untargeted metabolomics 13 .
We also report our efforts to establish a MS/MS spectral library of bioactive compounds obtained from medicinal plants that are used in traditional East Asian medicine. Although many phytochemicals have been previously found in medicinal plants, most of them vary in the historical collections of natural product chemistry laboratories and their MS/MS spectra have not been reported. Benchmarking recent efforts leading to the monoterpene indole alkaloid database (MIADB), and isoquinoline alkaloids and other annonaceous metabolites database (IQAMDB), which are spectral libraries built with compounds from historical collections of various natural product chemistry laboratories 14,15 , we established an MS/MS spectral library using 223 pure phytochemicals obtained from the legacy compound library of the Natural Products Research Institute, Seoul National University (SSK Legacy Library, named after Sam Sik Kang, who compiled the library over the course of 30 years). MS/MS spectra were acquired for all ionized molecules in positive (ESI+) and negative (ESI−) ion modes, which yielded 184 positive and 152 negative ion mode spectra. Compounds with low ionization efficiencies in each ionization mode were excluded. This spectral library will accelerate the annotation of phytochemicals in future metabolomic studies of medicinal plants. The chemical ontology of the phytochemicals included in the spectral library was estimated using the NPClassifier 16 and is summarized in Fig. 2.  www.nature.com/scientificdata www.nature.com/scientificdata/ National University), coupled with NMR spectra for structural identification. The compounds were dissolved in 50% aqueous MeOH at 100 μg/mL concentration for LC-MS/MS analysis.

LC-MS/MS data acquisition.
To compile the plant extract dataset, LC-MS/MS data were acquired using a Waters Acquity UPLC system (Waters Co., Milford, MA, USA) coupled to a Xevo G2 Q-TOF mass spectrometer (Waters MS Technologies, Manchester, UK) equipped with an electrospray ionization interface (ESI). To compose the spectral library, LC-MS/MS data were acquired using a Waters Acquity I-Class UPLC system linked to a Waters VION IMS Q-TOF MS equipped with an ESI interface. Chromatographic separation was performed using a Waters BEH C 18 column (50 × 2.1 mm, 1.7 μm), which was eluted with a mixture of water (solvent A) and Molecular networking analysis. Molecular networks were created using the classical molecular networking workflow on the GNPS web platform 10 . The networks were then created, where edges were filtered to have a cosine score above 0.65 and more than 5 matched peaks, with the precursor and fragment ion mass tolerance of 0.02 Da. The library spectra were searched to find annotation with the same score and matched peaks.
The molecular network with the positive ion mode data can be accessed via: https://gnps.ucsd.edu/ ProteoSAFe/status.jsp?task=94bd6547c84341ddaaff2e4599247871 The   www.nature.com/scientificdata www.nature.com/scientificdata/ to the standard compounds were exported as.mgf files. The metadata used to establish the spectral library, including SMILES and InChI identifiers of the structures, is provided as Supplementary Table 1.

technical Validation
Molecular networking-based overview of the KP337 dataset. Classical molecular networking analyses were performed with the datasets acquired in positive and negative ion mode 18 , respectively, to provide an overview of the specialized metabolome of the analyzed medicinal plants. The positive ion mode data yielded a molecular network consisting of 16,533 spectral nodes, while the negative ion mode data gave a network of 6,570 spectral nodes. The MolNetEnhancer workflow assigned class annotations to molecular families based on the spectral matching-based annotation 19 ; the resulting molecular networks are shown in Fig. 3, and the class annotations are summarized in Table 1. In both networks, phenylpropanoids and polyketides account for the largest portion of the annotated molecular networks. This may be due to the high polyphenol content of the medicinal www.nature.com/scientificdata www.nature.com/scientificdata/ plants. Organic nitrogen compounds and alkaloids and derivatives were only observed in the positive ion mode data-based network because of the basicity of nitrogen-containing compounds. For the positive and negative ion mode networks, 81.6% (13,488 of 16,533) and 76.2% (5,005 of 6,570) of the spectral nodes were respectively unannotated. This can seem to be too low, but it needs to be denoted that the class of each molecular family was annotated only based on the spectral matching. The number of publicly available reference spectra are still much lower than the number of known phytochemicals; thus, application of in silico spectral annotation methods would increase the annotation rate, as we demonstrated in a previous study on specialized metabolome of the family Rhamnaceae 20 .
Spectroscopic validation of the phytochemicals. The structures of the purified phytochemicals were confirmed by manual inspection of the nuclear magnetic resonance (NMR) and MS spectra.

Dereplication of the KP medicinal plants data against the spectral library. For further validation
of the spectral library, we re-established the molecular networks of the KP337 dataset 18 using the SSK legacy spectral library, together with all the spectral libraries available in GNPS. Consequently, 11 compounds (7 in positive and 4 in negative ion mode) were matched as the candidates with the highest scores ( Table 2). Most of the sample occurrences in the matched spectra correlated with the previous reports of the matched molecules from taxonomically same or close species, which supported the reliability of the dereplication result. γ-Fagarine was isolated from Phellodendron amurense 21 and Dictamnus albus 22 , while sesamin was found in Asarum heterotropoides 23 , and oxypeucedanin methanolate and pabulenol were isolated from Angelica dahurica 24,25 . Isolariciresinol, which was reported from Rubia argyi 26 , has not previously been reported in Patrinia scabiosifolia; however, it was reported from P. scabra, another species of the genus 27 . Due to the possible conservation of biosynthesis in close taxa,  www.nature.com/scientificdata www.nature.com/scientificdata/ species in the same genus or family often contain the same or similar specialized metabolites 28 . Thus, the occurrence of isolacriciresnol in P. scabiosifolia can be supported by the occurrence of the same compound in P. scabra. Spectral matching suggested the occurrence of trans-khellactone and peucedanol in Glehnia littoralis, but neither of these two compounds have been reported in the plant. However, six O-acyl derivatives of cis-khellactone were reported from G. littoralis 29 . Along with the spectral matching results, this suggests the occurrence of cis-khellactone in G. littoralis, as the cis/trans isomers cannot be distinguished by MS/MS analysis. Similarly, oxypeucedanol has not been reported in G. littoralis, but multiple oxypeucedanol glycosides were reported from G. littoralis, which supports the occurrence of the aglycone in this plant 30 . These cases simultaneously highlight the applicability and value of the dataset and spectral library introduced in this Data Descriptor; the coverage of spectral matching was expanded, and the occurrence of previously known compounds can be easily estimated by searching the spectra against the dataset.

Code availability
The data processing workflow for establishing the spectral library is available in GNPS.