The omics data set includes transcriptome sequencing, genomes, metagenomes, and chloroplast genomes. The bioinformation of the data sets was summarized, and a list was formed for easy searching. Germplasm data include species and cultivar information. Information was also collected on more than 500 compounds from 13 Panax species by searching the literature on phytochemistry between 1963 and 2021. The above information and that on chemical compositions were manually collected from publications, online resources, and other literature.

Genomic data sets primarily contain genome sequences, gene sequences, protein sequences, gff, gene functional annotations, and transcription factor (TF) annotations. Gene sequence and gff files of P. ginseng in the database were obtained from the ginseng genomic database (http://ginsengdb.snu.ac.kr/data.php). The 5th P. notoginseng genome assembly composed of 2.41 Gb from a previous study was included (Yang et al., 2021). Genome assemblies of P. ginseng and P. notoginseng included 59,352 and 47,870 genes, respectively. Genes were functionally annotated in InterPro (protein family; http://www.ebi.ac.uk/interpro/), pfam (protein family; http://pfam.xfam.org/), GO (Gene Ontology; http://geneontology.org/), NCBI non-redundant protein (Nr) database (http://www.ncbi.nlm.nih.gov), and Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg) by using BLASTX (e-value < 10⁻⁵; http://www.ncbi.nlm.nih.gov/BLAST/). The transcription factors were identified with PlantTFDB serve (http://planttfdb.gao-lab.org/).

RNA-Seq data sets of P. ginseng and P. notoginseng with biological duplication were selected. Data sets included different tissues, development times, and treatments. Filtered RNA-seq was mapped to the latest reference genome by HISAT2 (Zhang et al., 2021). Matched reads were subsequently presented to StringTie for de novo assembly, during which the expression level of each gene and the isoform was estimated (Pertea et al., 2015). Gene expression maps were prepared by manual selection on the Gene page.

FastQC software (v0.11.5) was used for quality control of raw reads from Illumina HiSeqTM2000 (WuHang, HuNan, China) sequencing, and clean reads were mapped to the latest reference genome using the BWA software (v0.7.15-R1140; Li and Durbin, 2009; De Sena Brandine and Smith, 2019). The GATK software (v4.0.9.0) was used for single nucleotide polymorphisms (SNPs) identification (Tajima, 1989). The ANNOVAR software was used to calculate the number of mutation types of SNP sites in each sample (Wang et al., 2010). The software LUMPY (v0.2.13) was used to detect structural variation (SV) in this project (Layer et al., 2014). PCA of the sample group structure was performed by using PLINK (v1.9b6) and EIGENSOFT (EIG-6.1.4) software (Patterson et al., 2006; Price et al., 2006; Purcell et al., 2007). The software ADMIXTURE (v1.3.0; Alexander et al., 2009) was used to analyze population structure. The software SNPhylo 2 (Lee et al., 2014) was used to construct evolutionary relations between different samples, and Phylogeny.IO (https://github.com/oist/phylogeny-io) was used to construct the phylogenetic tree. A graph of linkage disequilibrium was obtained by PopLDdecay (v3.40) analysis (Zhang et al., 2019). GCTA (v1.93.2) software was performed for GWAS analysis, and a mixed linear model (MLM) was adopted as a modeling method (Yang et al., 2011).

A total of 20.3 G of PacBio Iso-Seq data and 92.6 G of raw RNA-seq data of three representative tissues (roots, stems, leaves) were acquired from P. notoginseng sequencing. The RNA from the three tissues was mixed to establish two SMRT cells for SMRT sequencing. Smrtlink 5.0 software was used to process the raw data, and the obtained redundant sequences were clustered by an iterative clustering for error correction algorithm to obtain full-length consensus sequences (FL; Pertea et al., 2015). Quiver was used to align FL sequences and non-FL reads. The high-quality sequence with accuracy >0.99 obtained in the previous step was further aligned with the assembled data from Illumina RNA-seq by using Lordec (v0.9) and mapped to the reference genome by using GMAP software (Wu and Watanabe, 2005).

PanaxGDB is based on the Apache web server (http://www.apache.org), adopted ThinkPHP5.1 (http://www.thinkphP.cn)-based FastAdmin templates, included frameworks of Codelgniter (https://www.codeigniter.com/) and Bootstrap (https://getbootstraP.com), and applied programming languages including CSS, PHP-HTML5, and JavaScript. MySQL (https://www.mysql.com) was used for data sorting, storage, and management, and the AJAX asynchronous loading scheme was used for quick data loading and function implementation. To provide an interactive user experience, Echarts (https://echarts.apache.org/zh/index.html), JBrowse (http://jbrowse.org), Phylogeny.IO (https://github.com/oist/phylogeny-io), and JQuery (https://jquery.com) were applied. The File Transfer Protocol-based download function was also provided with a transfer speed of up to 50 Mbps. Additionally, all functions in PanaxGDB were tested for web browsing on Safari, Chrome, Firefox, IE, and Edge by mobile phone, pad, and computer.

PanaxGDB with a user-friendly web interface includes nine pages: Home, Species, Cultivars, Gene, Compounds, Population, Tools, and Download (Figure 1). The Home page presents an overview of PanaxGDB as well as links to popular ginseng sites and biological tools. Other pages represent the five main topics of PanaxGDB.

Two modules of the germplasm resources of Panax are included in PanaxGDB: Species (Figure 2A) and Cultivars (Figure 2B). These two pages can help users quickly understand the species structure of Panax and the breeding status of Panax cultivars, respectively. To determine species, authoritative plant taxonomic keys were used. Species not authenticated were identified according to morphological and molecular data in various public references, and species of uncertain status were indicated. Panax assamicus found in northeast India, which was not accepted by the Board of Trustees of the Royal Botanic Gardens, was cited as P. bipinnatifidus var. angustifolius. Similarly, Panax shangianus and Panax omeiensis were not included in the authoritative source for plant nomenclature. Although Panax vietnamensis var. fuscidiscus and P. vietnamensis var. langbianensis were not included in authoritative plant taxonomies, morphological and molecular evidence indicate they are new varieties (Van Duy et al., 2016).

Sixteen Panax species are summarized in the database, and eight categories of information are provided for each species on the Species page, including pictures, common name, synonyms, chromosome number, utilization, distribution, description, and references (Figure 2A). In China, local provincial authorities register and identify cultivars of medicinal plants, and as result, information on various cultivars is not concentrated. Therefore, enterprises and producers involved in Panax cultivation are limited in the collection of cultivar information, which is difficult to meet market demand. On the Cultivars page, those participating in the Panax industry can easily obtain the current breeding status of Panax cultivars to screen optimal cultivars for different products under local environmental conditions. The Cultivars page contains eight categories of information on cultivars, including crop variety number, uses, area of production, description, pictures, breeders or organizations, and references (Figure 2B). A total of fifty P. ginseng cultivars are recorded in PanaxGDB, including 19 from China, 30 from Korea, and one from Japan, as well as two cultivars of P. quinquefolius, one of P. japonicus, and nine of P. notoginseng from China.

PanaxGDB provides viewing of the latest assemblies of Panax genomes (P. ginseng and P. notoginseng). Genome assembly files of P. notoginseng are open to the public for research for the first time. For convenience and practicality, genome assemblies were re-annotated and uploaded. On the Gene page, users can enter a TF, GO, and KEGG orthology identifiers (IDs), Pfam domain, InterPro domain, or gene ID in different search boxes to obtain detailed corresponding information, such as structural and functional annotation of genes of interest. For each gene, 22 categories of information are provided, such as gene locus, length, and downloadable sequence, among others, which can be displayed using the “Operate” option in the upper-right of the display list (Figure 2C). Users can also access the ginseng genome databases from the Home page. In the gene expression module, one or more genes can be selected according to specific needs, and gene expression levels of different groups or treatments can be visualized from a heat map of ticked, related genes. Heat maps provide information on data accession numbers and associated reference (Figure 2C).

On the Tool page, JBrowse and Blast can help users visualize Panax genomes and sequence alignments. Currently, P. ginseng and P. notoginseng genomes, gene annotations, and information on Panax variants can be imported into JBrowse. Sets of genomes, gene and density information on SNP/300 kb, and allele frequency can be explored in PanaxGDB in JBrowse (Figure 2D). The Blast tool is provided for sequence alignment with Sequence coding for amino acids in protein (CDS) and protein sequences of P. ginseng and P. notoginseng and uses the input of a similar sequence (Figure 2E). The final Blast result can be downloaded in HTML format.

Secondary metabolites in Panax are pharmacologically active. Therefore, a metabolites page was provided with an interface (Compounds page) to retrieve, visualize, and investigate metabolite distribution in 12 Panax species (Figure 2F). The page includes 587 compounds of Panax, of which 492 are saponins. Damarane-type saponins accounted for 76.8% (451) of the known compounds. More than 200 saponins have been detected in P. ginseng and P. notoginseng according to the available phytochemical and pharmacological literature. In the Statistics modules, the proportion, quantity, and distribution of different types of compounds across different species can be visualized. In the list of chemical compounds, users can find ginsenoside Rb1, Rb2, Rc, Rd, Rb3, Re, Rg1, 20(S)-Rg2, Rh1, Rf, and Ro, notoginsenoside R1, and gypenoside XVII in at least eight Panax species. More than two Panax species contain 148 types of saponins. However, some characteristic saponins, such as quinquenoside I, II, III, IV, and V, are found only in P. quinquefolium (Yoshikawa et al., 1998). According to statistics, characteristic saponins are found in P. ginseng, P. notoginseng, P. quinquefolius, P. japonicus, P. japonicus var. major, P. vietnamensis, P. stipuleanatus, and P. bipinnatifidus. However, the list might be incomplete because of insufficient detection depth. To search for additional information, users can search by entering the name of a specific compound, compound type, or species origin. Users can click on a metabolite to link the metabolite information subpage and complete metabolite information, including biochemistry, biosynthesis pathway, and its species origin. Another hidden effect of this page is to help understand validated biosynthetic pathways of compounds, especially saponins, and validated genes in those pathways.

The resequencing database was set for Panax breeding and germplasm resource management. Data from P. notoginseng were collected on the Population page, and 200 3-year-old P. notoginseng samples from Yunnan, China, with nine important agricultural traits were randomly selected for resequencing. The average sequencing size of each sample was 28.83 Gb. Results were visualized and uploaded to the Population page with three modules which included Statistics (Figure 3A), Analysis (Figure 3B), and Variations (Figure 3C). Statistical results of the resequencing can be seen on the Statistics subpage. The Analysis subpage includes population structure, phylogenetic tree, decay of linkage disequilibrium, and Manhattan figure. All images and corresponding data on this page can be downloaded. Users can also search for SNPs on the Variations subpage based on results of phenotype evaluation and GWAS.

A total of 1,615,492,994 SNPs were obtained in 200 samples. There were 1,462,055,399 SNPs of 200 samples within intergenic regions, 114,939,464 within intronic regions, and 38,498,131 within exons. These data can be visualized on the Statistics subpage. On the Analysis subpage, the phylogenetic tree was derived from the SNP data, and the tree shows P. notoginseng of Wenshan and can be divided into five subgroups. In the GWAS module, there were 192 highly significant peaks on 12 pseudochromosomes associated with traits (root weight and leaf width) related to saponin yield. Those SNPs can be found on the Variations pages, and 22 occurred in CDSs (Supplementary Table 1), which can be found on the Gene page. Those genes were annotated as proteins associated with cell metabolism and growth, including peptidyl-prolyl cis-trans isomerase, aminotransferase, oxoglutarate dioxygenase, 2-isopropylmalate synthase, WD40-repeat protein, AMP-dependent synthetase, and phosphatidylinositol N-acetylglucosaminyltransferase, protein kinases, or as plant stress-related proteins, including heat shock protein (Supplementary Table 1).

To make it more convenient to use Panax omics data, all publicly available data were gathered and classified. The Download page contains information and links to all available public Panax omics data, and long-read sequencing and resequencing data were published for the first time in PanaxGDB (Figure 3D). Downloadable content in this module includes 405 transcriptome sequencings, two genomes, 21 metagenomes, and 67 chloroplast genomes. The visual data can help users to summarize information. Because the database is an open platform, prospective ginseng omics data are welcomed and can be included in the database by contacting the authors through the Data Submission page.