Chromosome-level genome of wild-simulated Panax ginseng identifies SNP markers for germplasm and medicinal quality evaluation

Graphical abstract

Highlights

• •Chromosome-level genome of wild-simulated ginseng.
• •Allotetraploid origin inferred from temporally proximate subgenome WGD events.
• •105 diagnostic SNPs enable germplasm authentication.
• •Genetic-chemical model standardizes medicinal quality control.

Abstract

Introduction

Panax ginseng C. A. Mey., a precious traditional medicinal herb, demonstrates diverse pharmacological activities, including immunomodulation and anti-fatigue effects. However, prolonged cultivation has led to germplasm admixture and cultivar degeneration, resulting in inconsistent quality that severely compromises its medicinal value and industrial standardization. Therefore, establishing accurate and efficient germplasm evaluation tools is critical for ensuring the quality of P. ginseng medicinal materials.

Objectives

To integrate whole-genome sequencing and chemical fingerprinting for ginseng germplasm identification and quality consistency assessment.

Methods

Chromosome-level genomes of ginseng from Jilin (JA) and Liaoning (FC) were assembled using PacBio HiFi, Illumina, and Hi-C technologies. Gene family identification and phylogenetic analysis were performed across 13 representative species. Using the JA genome as the reference, we constructed a pan-genome incorporating 7 ginseng genomes to dissect gene repertoire composition and structural variation distribution across ginseng populations. Population structure analysis with 76 individual ginseng samples revealed genetic diversity, and its integration with HPLC chemical fingerprints provided a joint assessment of quality consistency.

Results

We assembled two chromosome-level genomes and, through comparative genomics, revealed significant expansion of ginsenoside biosynthesis-related gene families and subgenome divergence. The core gene set accounted for 54.1% of the pan-genome, indicating high genetic conservation. SNP distribution patterns from population resequencing enabled the development of germplasm-specific molecular markers and a genetic-chemical integrated evaluation model.

Conclusion

The molecular marker system and genetic-chemical joint assessment developed here provide reliable novel tools for germplasm identification and quality control, advancing standardization in the ginseng industry.

Introduction

Panax ginseng C. A. Mey., a perennial medicinal plant of the Araliaceae family, is revered as the "King of Herbs" in traditional Chinese medicine for thousands of years [1], [2]. Recognized as one of the most valuable medicinal resources, ginseng has been traditionally used to tonify qi to prevent collapse, fortify the spleen and boost lung function, tranquilize the spirit and sharpen cognition, enrich blood and generate body fluids. Ginseng is widely prescribed for conditions such as qi deficiency, weakness, poor appetite due to spleen deficiency, and chronic cough from lung deficiency [3], [4], [5], [6], [7]. Modern pharmacological studies have further confirmed that ginsenosides, the primary bioactive components of ginseng, exhibit diverse therapeutic effects, including immunomodulation, anti-inflammatory activity, and antitumor properties. Currently, ginseng resources are primarily classified into three types: wild ginseng, wild-simulated ginseng, and cultivated ginseng. Wild ginseng grows entirely in natural forest environments without human intervention, while cultivated ginseng includes horticultural varieties such as common ginseng, Bian-tiao ginseng, and Shi-zhu ginseng. China has successfully bred over 20 registered cultivars, including "Jilin Huangguo" and "Xinkaihe No. 1". However, rising global demand for ginseng has led a conservation crisis, in which wild populations are collapsing due to deforestation [8], [9], habitat degradation, and over-exploitation, whereas cultivated germplasm is constrained by poorly characterized genetic backgrounds and erosion of genetic diversity, thereby undermining sustainable use and industrial development.

Despite nearly 2,000 years of cultivation history, systematic research on ginseng germplasm is still limited [2], [9]. The depletion of wild resources due to habitat degradation further complicates conservation and utilization efforts. Ginseng has a long growth cycle and possesses an allotetraploid genome (2n = 4x = 48) with high heterozygosity and poorly defined phenotypic characteristics [10], making genetic stability maintenance difficult in cultivation. Traditional breeding approaches depending on phenotypic selection remain constrained by the incomplete elucidation of genetic mechanisms governing key agronomic traits in ginseng, seriously limiting the development of superior cultivars and falling short of breeding efficiency requirements. Therefore, comprehensive studies on ginseng germplasm resources across broader geographical ranges are urgently needed. Such studies should employ population genetics and molecular breeding to establish accurate and efficient molecular evaluation systems. The systems are crucial for deciphering genetic stability, identifying functional genes, and facilitating targeted breeding, with substantial implications for both theoretical understanding and practical applications in ginseng research.

Recent advances in ginseng genomics include whole-genome sequencing and genetic maps construction for multiple cultivated varieties. However, high-quality reference genomes are still lacking for ginseng under wild-simulated cultivation, which exhibits superior medicinal properties. While traditional molecular markers including random amplified polymorphic DNA (RAPD), simple sequence repeat (SSR), and insertion-deletion (InDel) have established the basis for ginseng genetic studies [11], [12], [13], [14], their reliability and resolution capability are significantly constrained by the tetraploid genome complexity [15]. The rapid development of sequencing technologies has provided new research approaches for large-scale molecular marker screening. Genome-wide single nucleotide polymorphism (SNP) markers, with their advantages of broad genomic coverage, high polymorphism, and high-throughput detection capacity, have become essential tools in plant population genetics research and molecular breeding [16], [17], [18], [19], [20]. Platforms such as Fluidigm and Axiom arrays facilitate efficient genotyping and marker-assisted selection (MAS) [21], [22], [23], demonstrating success in crops including rice, wheat, and cucurbits [24], [25], [26], [27], [28]. These developments show strong potential for ginseng improvement. Furthermore, they provide essential genomic tools for studying ginseng cultivated under wild-simulated condition and enable integrated genetic-chemical quality assessment systems, crucial for evaluating medicinal properties.

This study presents the first chromosome-level genome assembly of ginseng cultivated under wild-simulated condition, elucidating evolutionary characteristics of this allotetraploid species. The analysis revealed lineage-specific expansion of terpenoid biosynthesis-related gene families and polyploidization evolutionary patterns in ginseng. By analyzing subgenome divergence timing and conserved synteny with P. quinquefolius, uncovered the genomic structural basis underlying the formation of genetic diversity in ginseng. Furthermore, through pan-genome construction, we identified a high proportion of core gene clusters reflecting genetic conservation, while providing a structural variation map for investigating germplasm admixture mechanisms. To address the genetic complexity caused by germplasm admixture, this study developed an evaluation model for assessing population genetic diversity and fingerprint stability, facilitating systematic analysis of genetic and chemical quality stability characteristics across different production regions, while also establishing core SNP-based molecular markers to differentiate breeding lines, registered cultivars, and wild-simulated populations. Our findings advance understanding of polyploid evolution in medicinal plants and establish molecular foundations for quality evaluation of geo-authentic medicinal materials.

Materials and methods

Plant materials and sequencing

Two wild-simulated ginseng accessions, JA and FC, were selected from Jilin and Liaoning provinces, respectively, for whole-genome de novo sequencing. Additionally, 76 ginseng samples collected from eight different geographical regions in China were subjected to whole-genome resequencing (Table S1). Fresh leaves were flash-frozen in liquid nitrogen and stored at −80°C prior to DNA extraction. Genomic DNA was isolated from leaf tissues using a modified 3 × cetyltrimethylammonium bromide (CTAB) method [29], and DNA concentration was quantified by measuring A260 absorbance using an ND-2000 spectrophotometer [30]. For de novo sequencing, paired-end (PE150) libraries were constructed for Jilin (JA) and Liaoning (FC) following Illumina’s standard protocols and sequenced on the HiSeq X Ten platform. Additionally, 10-kb insert-size HiFi libraries were prepared for both JA and FC and sequenced on the PacBio Sequel II system for high-quality genome assembly. Hi-C library was also generated for JA and sequenced on the HiSeq 2000 platform to facilitate chromosome-scale scaffolding. For the 76 resequenced samples, PE libraries with insert sizes of 300–500 bp were constructed and sequenced on the MGI DNBSEQ-T7 platform to obtain whole-genome resequencing analysis data.

Genome assembly

The de novo assembly was performed using hifiasm v0.24.0 [31] with PacBio CCS reads to generate contig-level genome sequences. To further refine the primary assembly and reduce haplotype redundancy, we applied Purge_dups v1.2.6 [32] to eliminate heterozygous duplications prior to polishing and scaffolding. The assembled contigs underwent error correction through an initial polishing step with Racon v1.4.16 [33] using HiFi reads, followed by a secondary polishing step with Pilon v1.2 [34] using Illumina short reads. Assembly quality was evaluated based on Illumina read mapping rates and Benchmarking Universal Single-Copy Orthologs (BUSCO) completeness assessment. To achieve chromosome-scale scaffolding, Hi-C sequencing reads were aligned to the assembled scaffolds using Juicer v1.5 [35] to construct Hi-C interaction matrix. The resulting assembly was manually curated and refined through redundancy removal using Juicebox v1.11, yielding a chromosome-level genome assembly. And the assembled ginseng genome was partitioned into subgenomes A and B using SubPhaser [36].

Identification and annotation of repetitive sequences

Repetitive elements were identified and annotated through a combination of de novo identification and homology-based prediction approaches. De novo repeat identification was performed using RepeatModeler v1.0.8 [37] to detect potential repetitive elements in the genome. The resulting repeats were then combined with known repetitive sequences from the RepBase database [38] to construct a comprehensive repeat library. The integrated repeat library was subsequently analyzed with RepeatMasker v4.0.7 [39] to identify and annotate homologous sequences, thereby systematically characterizing genomic regions sharing similarity with known repetitive elements. To enhance the annotation of LTR retrotransposons (LTR-RTs), we first performed initial detection using LTRharvest [40] to comprehensively identify these biologically significant repetitive elements. Further refined analysis, including calculation of the LTR assembly index (LAI) and estimation of insertion times was conducted using LTR-Retriever.

Evolutionary analyses of LTR-RTs

To investigate the phylogenetic relationships among LTR-RTs in the subgenomes A and B of P. ginseng, subgenomes A and B of P. quinquefolius, P. notoginseng, and Aralia elata genomes, conserved motif combinations within the reverse transcriptase (RT) domains were first identified using MEME v4.11 [41]. Subsequently, multiple sequence alignments of Gypsy and Copia RT domain amino acid sequences were performed using the fftns module in MAFFT v7.305b [42]. Poorly aligned regions in these alignments were then trimmed using TrimAl v1.4.rev22 [43]. The maximum likelihood (ML) phylogenetic trees were constructed using RAxML 8.2.12 [44].

Gene annotation

For comprehensive gene annotation, we employed an integrated strategy incorporating ab initio prediction, homology-based prediction, and transcriptome evidence. The genome was initially trained using autoAugTrain.pl, followed by gene prediction with Augustus. Homology-based predictions were generated through protein sequence alignment using BLAST v2.9.0 and GeneWise v2.2.0. Transcriptomic evidence was derived from publicly available P. ginseng RNA-seq data [45], with stringent filtering criteria requiring ≥30% transcript coverage and >97% sequence identity. These diverse prediction sets were integrated using MAKER v2.31.10, with final annotation quality assessed through BUSCO v5.2.2 against the embryophyta_odb10 database to ensure completeness and reliability of the gene set.

Phylogenetic analyses

Orthologous and paralogous gene clusters were identified in 13 plant species, including Oryza sativa, Arabidopsis thaliana, Vitis vinifera, Coffea arabica, Artemisia annua, Daucus carota, Eleutherococcus senticosus, A. elata, P. quinquefolius, P. vietnamensis var. fuscidiscus, P. notoginseng, P. japonicus, and P. ginseng, using OrthoFinder v2.5.5 [46]. Single-copy gene sequences were aligned using MUSCLE, and all alignment results were merged to generate a super alignment matrix. To estimate species divergence times, MCMCtree from PAML v4.9 [47] was applied, with fossil calibration times obtained from the TimeTree database (https://www.timetree.org/) [48]. Additionally, CAFÉ [49] was employed to analyze expansions and contractions of orthologous gene families, further exploring their dynamic evolutionary changes across species.

Syntenic analyses

Homologous genes among species were identified using BLASTp v2.12.0, followed by syntenic block detection and collinearity analysis with MCscanX [50]. Whole-genome duplication (WGD) events in P. ginseng were determined based on the distribution of synonymous substitution rates (Ks). Orthologous gene pairs identified through synteny analysis were further analyzed using KaKs Calculator [51] to compute and compare Ks values for both intra-species paralogs and inter-species orthologs, enabling the estimation of WGD event timing.

Pan-genome analysis of P. Ginseng

A pan-genome was constructed using seven representative P. ginseng genomes consisting of four newly sequenced genotypes (JA1 and JA2 from Jilin, FC1 and FC2 from Liaoning) along with three published cultivated genomes including two Chinese cultivars NGJ and JTT [10], [52] and one Korean variety TYP [53]. For protein-coding genes, only the longest transcript per locus was retained. Homologous gene families were clustered using OrthoFinder [46] (parameters: –S diamond –og). Based on their distribution across genomes, gene clusters were categorized into: the core genome (shared by all seven genomes), the dispensable genome (present in 2–6 genomes), and the private genome (genes unique to a single genome or unclustered) [54].

Variation identification based on comparative analysis among ginseng varieties

Genome-wide variant detection, including SNP, InDel, and structural variant (SV) identification, was performed using the JA1 genome as a reference compared to six other ginseng genomes. Sequence alignment was conducted using nucmer, followed by filtering with delta-filter, and alignment orientation and ID standardization using chroder. Secondary variant calling was then performed, with SNPs and InDels identified using delta2vcf. For structural variant analysis, SyRI [55] was employed to detect insertions (INS), deletions (DEL), inversions (INV), translocations (TRANS), and copy number variations (CNS) [56].

SNP calling and filtering

Raw whole-genome resequencing data from 76 ginseng samples were first preprocessed for quality control using Trimmomatic. The reads were then aligned to the P. ginseng reference genome using BWA-MEM [57], followed by sorting and filtering for uniquely mapped reads with SAMTools. After data filtering and read group processing with Picard v2.2.1, SNP calling was performed using DeepVariant v1.6.0 [58].

Variants were filtered using bcftools v1.11 [59] to ensure reliability for population structure and diversity analyses: (i) low-quality SNPs (<Q30) were removed, (ii) SNPs with sequencing depth < 3 were filtered out, (iii) only SNPs present in at least 90% of samples were retained, and (iv) SNPs with minor allele frequency (MAF) < 0.03 were excluded.

For SNP chip development across different P. ginseng strains, additional filtering steps were applied: SNPs located within 50 bp of another SNP were excluded to minimize interference from neighboring variants. To ensure sufficient genetic diversity, selected SNPs included both heterozygous and at least one homozygous genotype. Furthermore, to enhance accuracy, only SNPs with unique 50 bp flanking sequences in the P. ginseng genome were retained. After filtering, a final set of 105 high-confidence SNPs was obtained.

Population structure and genetic diversity analysis

To visualize genetic relationships among ginseng samples, discriminant analysis of principal components (DAPC) was performed using the filtered SNPs. Population structure was assessed with Admixture v1.3.0 [60], Testing potential cluster numbers (K = 2–15) with 1,000 replicates. The optimal K value was determined based on cross-validation error. A ML phylogenetic tree was constructed using RAxML v8.2.11 [44] and visualized with iTOL (https://itol.embl.de/).

Fixation index (Fst) and nucleotide diversity (π) were calculated using vcftools v0.1.16. Observed heterozygosity (Ho) and expected heterozygosity (He) were computed with Plink. To evaluate genetic stability, allele frequency variance and Shannon entropy were analyzed. The mean polymorphism information content (PIC) was used to quantify population diversity, calculated as:

$$\overline{\mathrm{P}\mathrm{I}\mathrm{C}}=\frac{1}{N}\sum _{i=1}^N{\mathit{PIC}}_i$$ (1)

The normalized Shannon-Weiner diversity index, representing allele frequency uniformity and stability, was derived as:

$$\begin{array}{c}{H}_{mean}=\frac{{\sum }_{i=1}^N\left[-\left(p_{i}log{p}_i+q_{i}log{q}_i\right)\right]}{N}\end{array}$$ (2)

here, $p_i$ and $q_i=1-p_i$ denote the frequencies of the i-th allele.

HPLC analysis of ginsenosides and related compounds

The composition of ginseng samples (Table S2) was analyzed using high-performance liquid chromatography (HPLC) (Agilent 1260 Infinity II system, Agilent Technologies Inc.). The reference standard solution was prepared as a 0.2 mg/mL methanol mixture containing Rg1, Re, Rb1, Rc, Ro, and Rd. The analytical conditions were as follows: Agilent InfinityLab Poroshell HPH column (3.0 × 150 mm, 2.7 μm); mobile phase consisting of acetonitrile (A) and 0.1% phosphoric acid aqueous solution (B); gradient elution program: 0–12 min 19% A, 12–14 min 19%-30% A, 14–22 min 30% A, 22–30 min 30%–33% A, 30–40 min 33%-95% A; flow rate 0.8 mL/min; column temperature 30°C; detection wavelength 203 nm; injection volume 10 μL.

Statistical assessment of HWEP and integrative genetic chemical analysis

To evaluate the chemical stability of ginseng samples and determine the required sample size for equilibrium fluctuation analysis, a chemical equilibrium population model was established based on the principle of Shannon entropy. The specific formula is as follows:

$$\begin{array}{c}{S}_p=-{\sum }_{i=1}^n{p}_{i1}ln{p}_{i2}\end{array}$$ (3)

where $S_p$ represents the total information entropy of the fingerprint, $p_{i1}$ denotes the response value (peak area) of a specific characteristic peak in the fingerprint of an individual medicinal material within a genetic equilibrium population, and $p_{i2}$ is the percentage of the corresponding peak’s response value relative to the entire fingerprint. The formula for the equilibrium population size is:

$$\begin{array}{c}n\ge {\left(K\bullet RSD\right)}^2\end{array}$$ (4)

here, K is the preparation control coefficient, set at K = 25.6, yielding the Hardy-Weinberg equilibrium population number. Using $H_{\mathit{mean}}$ and the Hardy-Weinberg equilibrium population number derived from fingerprint data, analysis of variance (ANOVA) and correlation analysis were performed to assess genetic homogeneity and chemical stability. To accurately assess the intrinsic relationship between genomic diversity and chemical traits while controlling for confounding environmental factors and cultivation age, we implemented an environmental correction protocol. Core environmental variables (Precipitation of Warmest Quarter, Temperature Annual Range, and Elevation) contributing most significantly to habitat suitability (AUC = 0.995) were selected. These factors, combined with cultivation age, were used as covariates in subsequent genetic-chemical correlation analyses, thereby constructing an integrated genetic-chemical evaluation system.

Results

Assembly, annotation and characteristics of two ginseng genomes under wild-simulated cultivation

We performed whole-genome sequencing of wild-simulated ginseng (JA from Jilin and FC from Liaoning), two major production regions in China. The JA sample generated 90.41 Gb of Illumina paired-end reads, 47.48 Gb of PacBio high-fidelity (HiFi) reads (N50 = 19.54 kb), and 293.46 Gb of Hi-C data for de novo assembly. Similarly, FC was subjected to sequencing involving 40.47 Gb of HiFi reads (N50 19.37 kb), and 93.55 Gb of Illumina paired-end reads (Fig. 1, Table 1). We used these data for genome assembly. The final assembly of the JA genome was 3.48 Gb, with an N50 size of 146.58 Mb. Among these assembled sequences, 98.6% were anchored onto 24 chromosomes ranging in size from 89.6 Mb to 194.8 Mb (Fig. 1, Table 1). The assembled FC genome had a size of 3.46 Gb with an N50 size of 145.93 Mb, where 98.3% of sequences were anchored to 24 chromosomes with an average chromosome length of 141.64 Mb. To evaluate assembly quality, we performed BUSCO assessments for both JA and FC genomes. The results showed 98.8% completeness for JA (S: 9.5%, D: 89.3%, F: 0.9%, M: 0.3%) and 98.9% completeness for FC (S: 13.5%, D: 85.4%, F: 0.9%, M: 0.2%).

Fig. 1.

An overview of the genomic features of JA and FC. The outermost circle represents the chromosomes: the 24 pseudochromosomes of JA (left, counterclockwise) and the 24 pseudochromosomes of FC (right, clockwise). (a) Guanine-cytosine (GC) content distribution. (b) Density of Gypsy-type retrotransposons. (c) Density of Copia-type retrotransposons. (d) Distribution of full LTR elements. (e) LAI index distribution. (f) Gene density distribution. The inner circle illustrates the syntenic relationships between JA and FC genomes.

Table 1.

Summary of JA and FC genome assembly and annotation.

Sequencing	JA	FC
Raw bases from WGS-Illumina (Gb)	90.41	93.55
HiFi (Gb)	47.48	40.47
Raw bases from Hi-C (Gb)	293.46	NA
Assembly
Number of contigs	1020	2913
Assembled genome size (Gb)	3.48	3.46
Contig N50 (Mb)	12.62	3.37
Scaffold N50 (Mb)	146.58	145.93
Average Scaffold length (bp)	10,790,496	59,619,439
GC content (%)	34.76	34.74
Anchorage to chromosomes (%)	98.6	98.3
Complete BUSCOs (%)	98.8	98.9
Annotation
Percentage of repeat sequences (%)	87.30	87.28
LTR rate (%)	80.77	81.65
Number of predicted genes	76,934	76,377
Average gene length (bp)	5,603	5,551
Average CDS length (bp)	1,069	1,066

Through ab initio prediction, homology-based annotation, and transcriptome data, we predicted 76,934 and 76,377 protein-coding genes in JA and FC, respectively. The BUSCO completeness scores of JA and FC were 98.4% (S:16.4%, D:82.0%, F:0.7%, M:0.9%) and 98.2% (S:18.6%, D:79.6%, F:0.9%, M:0.9%), respectively. These quality results validate the high continuity and accuracy of the P. ginseng genome assemblies and annotations.

Evolutionary insights from the phased subgenomes A and B of P. ginseng

We annotated repetitive sequences in JA using RepeatModeler, RepeatMasker, and LTRharvest, identifying a total of 3,038 Mb of repetitive sequences, accounting for 87.3% of the entire genome. Among these, LTR retrotransposons were the most abundant transposable elements, comprising 80.77% of the total genome. Within the LTR elements, the Gypsy and Copia accounted for 56.02% and 7.32% of the genome, respectively (Table S3).

To investigate the evolutionary dynamics of LTR-RTs in P. ginseng, we performed a systematic comparative analysis of repeat composition among P. ginseng, P. quinquefolius, P. notoginseng, and A. elata (Fig. 2(a)). Comparative analysis revealed significant enrichment of the Tekay and Athila (Gypsy superfamily) and the Angela and SIRE (Copia superfamily) in the ginseng genome relative to the other species (Fig. 2(a, b)). And the lineage-specific amplification pattern was further corroborated by phylogenetic analysis.

Fig. 2.

Evolutionary dynamics of TEs and subgenome characteristics in P. ginseng. (a) Proportions of the Gypsy and Copia subfamilies of LTR-RTs in twelve gymnosperms. (b) Phylogenetic relationship of the gypsy (left) and copia (right) subfamilies in P. ginseng. (c) The insertion time (MYAya) of dominant subgroups in four species. (d) Phylogenetic tree and gene family expansion/contraction of P. ginseng and related species. Orange indicates expansion, gray indicates contraction, orange circles and stars mark WGD and WGT events, respectively. (e) Ks distribution of subgenome A and B in P. ginseng, showing three polyploidization events: Pg-α, Pg-β, and γ triplication. (f) Chromosome collinearity between P. ginseng and P. quinquefolius subgenomes.

Based on the count and insertion time of their full-length LTR-RTs, the 24 chromosomes of JA were resolved into two distinct subgenomes (Fig. S1). The larger subgenome A (1.96 Gb) exhibited higher LTR-RT density, earlier insertion times, and elevated GC content (35.49%). Conversely, the smaller subgenome B (1.52 Gb) displayed lower LTR-RT abundance, more recent insertions, and reduced GC content (35.26%, p < 0.001). Further analysis of Gypsy and Copia insertion time across subgenomes revealed that Gypsy elements underwent large-scale accumulation primarily between approximately 0.21–1.4 Mya, with a burst of amplification occurring during the 0.21–0.22 Mya. The major insertion events for Copia elements were concentrated between 0.42–0.98 Mya (Fig. 2(c)). This LTR-RT accumulation profile was validated in the P. quinquefolius genome. Collectively, these findings provide crucial molecular evidence for the genomic evolutionary history of Panax species.

To further trace the origin of the allotetraploid P. ginseng, we conducted comparative genomic analyses using 13 plant species from Poaceae, Brassicaceae, Vitaceae, Rubiaceae, Asteraceae, Apiaceae, and Araliaceae, including the Subgenomes A and B of P. ginseng and P. quinquefolius (Fig. 2(d)). A total of 476 single-copy gene families were selected to construct a ML phylogenetic tree. The results revealed that Subgenome A and B of P. ginseng were most closely related to Subgenome A and B of P. quinquefolius, respectively, with an estimated divergence time of approximately 6.44 Mya. Their common ancestor diverged from A. elata around 13.83 Mya. Furthermore, extensive gene family expansions and contractions were observed in the Subgenomes A and B of P. ginseng. Subgenome A contained 1,923 expanded and 2,981 contracted gene families, while Subgenome B exhibited 1,357 expanded and 3,254 contracted gene families (Table S5). Functional enrichment analysis demonstrated that the expanded gene families in both subgenomes were significantly enriched in Gene Ontology (GO) terms related to "sesquiterpenoid biosynthetic process", "triterpenoid biosynthetic process", and "terpenoid biosynthetic process". These findings suggest that the high ginsenoside-related metabolic activity in P. ginseng is closely associated with the biosynthesis of sesquiterpenoids, triterpenoids, and other terpenoids.

Determining whether WGD events occurred simultaneously in Subgenome A and B is essential for understanding the genome evolution of P. ginseng. We analyzed the Ks density distribution and fourfold degenerate synonymous site (4DTv) of syntenic gene pairs both between Subgenome A and B and within each subgenome. The results showed nearly identical Ks and 4DTv distributions between Subgenome A and B, indicating that the two subgenomes likely underwent WGD events simultaneously. Further examination of the P. ginseng genome Ks revealed three polyploidization events (Fig. 2(e)). A distinct peak at 0.03 between Subgenome A and B suggests a Panax-specific WGD events (Pg-α duplication) to their divergence. Additionally, homologous gene pairs within Subgenome A (0.38–0.40) and Subgenome B (2.61–2.83) exhibited clear peaks, corresponding to the Panax-specific WGD (Pg-β duplication) and core-eudicot γ triplication, respectively. Chromosomal synteny analysis revealed extensive syntenic regions between Subgenome A and B of P. ginseng, showing a highly similar differentiation pattern to that observed between Subgenomes A and B of P. quinquefolius (Fig. 2(f)). This synteny further supports that the two subgenomes of P. ginseng experienced nearly synchronous WGD events before their divergence. Combined with the evolutionary relationship with P. quinquefolius, we propose that both P. ginseng and P. quinquefolius formed allopolyploid through hybridization following the same recent polyploidization event, with both polyploidization and geographic isolation jointly driving the speciation of Panax species.

Identification of variations in representative wild-simulated and cultivated ginseng

To investigate the genetic diversity of P. ginseng, we selected wild-simulated ginseng samples from Liaoning and Jilin provinces (FC and JA) and three published cultivated ginseng genomes as materials for genetic variant identification. Using the high-quality JA genome as the reference, we successfully resolved variations between its two haplotypes, HA and HB. Analysis of the two haplotypes in JA revealed 151,815 SNPs between them. Chr12A showed the highest SNP density, with 60,200 SNPs (396 SNPs/Mb), while Chr06B had the lowest density, with only 2,271 SNPs (27 SNPs/Mb) (Fig. 3(a)). The remaining chromosomes showed intermediate SNP distributions without extreme variations. Further examination of SNP distribution across chromosomes indicated that most SNPs were located in intergenic regions and introns (Fig. S2). However, two chromosomes, Chr07B and Chr08A, displayed a significantly higher proportion of SNPs in exonic regions compared to other chromosomes, suggesting distinct genetic variation patterns in their functional gene regions.

Fig. 3.

Genetic variations and pan-genome analysis of ginseng. (a) Manhattan plot of SNP density differences between the two haplotypes of the JA ginseng genome. The x-axis represents the chromosomes, and the y-axis represents the SNP density per Mb. (b) Gene family composition in the pan-genome of P. ginseng. The pie chart shows the proportions of each type of gene family, with blue, green, and red squares for core, dispensable, and private genomes, respectively. (c) SVs in the pan-genome of P. ginseng. The stacked bar chart shows the number of different types of SVs in each individual, with colors indicating the types.

Using the identified JA haplotype variations, we integrated the haplotype from FC and three published P. ginseng genomes to construct a pan-genome. By comparing 515,552 protein sequences from seven ginseng genomes, we mapped the distribution curves of core and species-specific gene families (Fig. 3(b)). In the ginseng pan-genome, 54.1% (35,260) were core genome, 44.2% (28,817) were dispensable genome, and only 1.7% (1,123) were private genome. Compared with cabbage (11.41%), maize (27.1%), rice (30%), and soybean (35.9%), ginseng showed a higher proportion of core gene families. To verify the robustness of this observation, we performed a focused pan-genome analysis exclusively on six Chinese ginseng genomes (excluding Korean sample TYP). This control experiment revealed a nearly identical core genome proportion (54.3%, Fig. S3(a)), indicating that P. ginseng exhibits relatively low genetic diversity in its genomic structure. This further suggests that, compared to other crops, the ginseng pan-genome is more stable in genetic variation and likely has a more uniform genetic background.

To elucidate evolutionary characteristics of the three genome types, a phylogenetic tree was constructed based on 21,997 single-copy gene. The results demonstrated that wild-simulated ginseng (JA) formed an independent branch, showing clear separation from all cultivated ginseng and another wild-simulated ginseng group (LC). Moreover, within another branch, the Korean cultivated sample TYP exhibited significant divergence from other samples (Fig. S3(b)). Functional enrichment analysis revealed that the core genome exhibited functional conservation in fundamental processes including terpenoid biosynthesis, signal transduction processes, and core components of genetic information processing (Fig. S3(c, f)), while the dispensable genome showed adaptive enrichment in environmental sensing and stress-response pathways (Fig. S3(d, g)). The private genome was associated with sample-specific processes such as vesicle transport (Fig. S3(e)).

Genomic variations, particularly short-sequence variations (such as SNPs and small-fragment insertions/deletions) and SVs, play crucial roles in adaptive evolution and diversity formation of species. To further investigate the variation characteristics of the ginseng genome, we conducted a comprehensive variation analysis and identified a total of 7,017,782 SNPs and 2,100,483 InDels, collectively referred to as short variations in this study. Additionally, we detected 26,125 SVs (Fig. 3(c)), including 3,782 INSs, 5,351 DELs, 254 inversions INVs, 1,358 translocations TRANs, and 15,380 CNVs. These variations collectively accounted for 4.20% of the total genome length, with certain differences observed in the quantity and types of SVs among different samples.

Ginseng population structure and genetic diversity among different cultivars

We performed WGS on 76 individual ginseng samples representing 3 different strains from 8 production areas in Jilin and Liaoning, China, generating a total of 2.25 Tb of data (Fig. 4(a)). Based on sample alignment results, we identified 39,294,162 SNPs and 7,980,301 InDels. Approximately 72.69% of SNPs were located in intergenic regions, 6.53% in intronic regions, and 8.39% in coding regions. We randomly selected 5,376 SNP loci and characterized the genetic relationships among all ginseng samples using ML phylogenetic trees, population structure, and DAPC analyses. All analyses indicated potential genetic admixture in ginseng germplasm resources with indistinct population differentiation (Fig. S4(c)). Individuals from breeding lines, cultivars, and wild-simulated ginseng were intermixed within groups without clear pattern classification (Fig. 4(b, c); Fig. S4(f)). Breeding lines (Population I, n = 52) included cultivated ginseng from Jilin, Liaoning and other regions; Population II represented registered cultivars (n = 15), mainly comprising improved cultivars developed through modern breeding techniques such as "Fuxing No.1", "Fuxing No.2" and "Xinkaihe No.1"; Population III represented wild-simulated ginseng (n = 9). Regarding population polymorphism levels, registered cultivars and wild-simulated ginseng showed higher π values of 0.177 and 0.175, respectively, while other breeding lines exhibited slightly lower polymorphism (Fig. S4(e)). Furthermore, π value of registered cultivars showed significant differences (p < 0.05) compared with both wild-simulated ginseng and breeding lines (Fig. 4(d)). The highest average Fst value of 0.028 was observed between registered cultivars and wild-simulated ginseng, indicating the presence of multiple shared genotypes within populations and suggesting limited genetic differentiation.

Fig. 4.

Genetic diversity and population structure of ginseng cultivars. (a) Geographic distribution of 76 ginseng samples. Map showing the geographic distribution of 76 ginseng samples from three cultivars across eight locations in Jilin and Liaoning, China. (b) Bayesian model-based clustering analysis with different number of groups (K = 2–6). The x-axis represents the number of groups (K), and each color represents one putative ancestral background. The y-axis quantifies the ancestry membership. (c) ML phylogenetic tree of ginseng samples. (d) Violin plot of π levels among ginseng groups. Colors correspond to each group. (e) DAPC analysis of three groups using 105 diagnostic SNPs, with colors designating group identity.

To characterize genetic differentiation characteristics among breeding lines, registered cultivars, and wild-simulated ginseng, the identified SNPs were filtered, yielding 105 unique, single-copy, and representative SNP loci. The results demonstrated that wild-simulated ginseng formed a distinct cluster on the discriminant function axis, while registered cultivars and breeding lines, despite partial overlap, exhibited significant separation in their principal component distribution trends (Fig. 4(e)). The Fst further confirmed that the selected SNPs effectively distinguished wild-simulated ginseng from the two cultivated types (Fig. S4(g)). This indicates that locus screening successfully captured genomic regions associated with population differentiation, providing candidate markers for the development of strain-specific molecular markers.

Joint chemical and genetic assessment of ginseng quality across different origins

To assess the genetic stability and chemical quality consistency of ginseng across different production regions, we analyzed the genetic diversity of 76 ginseng samples from eight regions in Jilin and Liaoning, China, including π value, Ho, He, and MAF. Nucleotide polymorphism analysis showed FSWL had the highest nucleotide diversity with a π value of 0.45, while JA, exhibited the lowest π value of 0.37. The remaining six regions displayed intermediate π values ranging from 0.38 to 0.40 (Table S6). Violin plots of π values (Fig. 5(a)) revealed broader distribution tails at higher π values for FSWL and HRB, with rapid convergence at lower π values. Other regions such as FC, FSX, and JA showed significantly different π distributions (p < 0.05). This trend was consistent with heterozygosity and MAF results. FSWL had the highest Ho of 0.65, He of 0.36, and MAF of 0.32, while KD, displayed the lowest genetic diversity (Fig. 5(b)). The mean value of PIC and H_mean further supported these findings, confirming FSWL's higher genetic diversity and more complex population structure (Fig. 5(c, d)). The Fst analysis indicated weak genetic divergence among most regions, except for FSWL, suggesting frequent gene flow or lack of significant genetic isolation.

Fig. 5.

Joint assessment of genetic diversity and chemical stability in ginseng across geographical origins. (a) Violin plots of nucleotide diversity (π) distribution among regions. (b) The proportion of homozygous alleles were decoded as green (minor type) and pink color (major type) and heterozygous alleles were decoded as gray color. Histogram of t of PIC (c) and H_mean (d) among ginseng from different regions. (e) HPLC fingerprint profiling of ginseng samples from 8 regions. Peaks marked with numbers correspond to: 1. Ginsenoside Rg₁; 2. Ginsenoside Re; 3. Ginsenoside Rb₁; 4. Ginsenoside Rc; 5. Ginsenoside Ro; 6. Ginsenoside Rd. (f) Correlation between genetic and chemical fluctuation of ginseng samples.

A comprehensive evaluation of traditional Chinese medicine quality stability requires considering both genetic background consistency and the quality stability of key bioactive components. HPLC fingerprint analysis at 203 nm identified 23 common peaks, including six characteristic peaks that were confirmed as ginsenosides Rg₁, Re, Rb₁, Rc, Ro, and Rd through comparison with reference standards (Fig. 5(e)). Intra-region similarity exceeded 0.9 for most regions, but FSWL and HRB exhibited lower similarity, reflecting greater chemical variability. Notably, FSWL showed the highest peak intensity variability with a relative standard deviation (RSD) of 32.94%, requiring a larger balanced sample size (n = 71.11) for reliable characterization, whereas KD had the lowest RSD of 7.17% and minimal sample size requirement (n = 3.37) (Table S7).

After controlling for confounding effects of cultivation age and core environmental factors (Fig. S5, Table S8 and Table S9), a correlation model between genetic uniformity (H_mean) and chemical stability (S_p) revealed a significant positive relationship (ρ = 0.965, p < 0.05). Regions with higher genetic diversity (e.g., FSWL) displayed greater chemical variability, while genetically stable regions (e.g., KD) showed superior chemical consistency (Fig. 5(f)). These results confirm that genomic polymorphisms directly drive chemical trait variation in P. ginseng. These findings provide novel insights into the formation of geo-authentic traits in medicinal plants. Long-term artificial selection may reduce genetic diversity while enhancing secondary metabolite stability. This study establishes a link between genetic diversity and chemical quality in ginseng, providing scientific basis for selecting optimal cultivation sites and breeding strategies aligned with good agricultural practices (GAP) for medicinal plants.

Discussion

A central challenge in medicinal plant genomics is the high-quality assemblies of highly heterozygous genomes. While numerous plant genomes have been successfully sequenced [61], [62], [63], [64], [65], [66], [67], [68], [69], assembling complex medicinal species such as P. ginseng, which exhibits high repeat content and polyploidy, remains particularly challenging. Xu et al. [70] first generated a draft ginseng genome using Illumina short-read sequencing, but the limitations of short-read technology restricted assembly continuity. Later, Li et al. [10] achieved a complete telomere-to-telomere (T2T) gap-free assembly through integration of long-read, short-read, and Hi-C data. However, current studies have focused on cultivated varieties, leaving wild-simulated ginseng germplasms with distinct medicinal properties underrepresented in genomic research. In this study, we employed a multi-platform sequencing approach combining PacBio HiFi, Illumina, and Hi-C technologies to assemble chromosome-level genomes of ginseng under wild-simulated cultivations from Jilin and Liaoning provinces, the primary production regions. Interestingly, despite using only 35.04% of the sequencing data volume compared to Li et al. [10], our JA assembly attained a contig N50 (146.58 Mb) nearly equivalent to their T2T genome (147.37 Mb). Gene annotation identified 76,934 and 76,377 protein-coding genes in the JA and FC genomes, respectively. These values exceed those reported by Xu et al. [70] (42,006 genes) and Yang et al. [71] (59,352 genes), while remaining comparable to Li et al.’s findings [10]. This advancement demonstrates that optimized sequencing strategies can produce high-quality assemblies with reduced data input, offering a cost-efficient solution for medicinal plant genomic studies.

Genomic structural analysis revealed that LTR-RTs accounted for 80.77% and 81.65% of the JA and FC genomes, respectively, significantly higher than the previously reported 62% [70] and 79.52% [71] in cultivated ginseng genomes. This discrepancy likely reflects both the improved resolution of long-read sequencing technology for complex repetitive regions and indicate potential adaptive changes in genomic architecture specific to wild-simulated ginseng. k-mer and PCA analyses clearly partitioned the JA genome into two subgenomes, with subgenome B showing a significantly lower LTR insertion frequency than subgenome A, and exhibiting later LTR insertion times than subgenome A. Furthermore, based on comparative analysis of repetitive element composition in related species, calculation of LTR-RT insertion time, collinearity analysis, and examination of published LTR data in Panax species, this pattern is inferred to be an intrinsic feature of allopolyploid evolution in P. ginseng potentially associated with subgenome differentiation patterns. The divergent LTR architecture demonstrates a discernible association with established patterns of allopolyploid evolution, with its synergies in spatiotemporal dynamics providing new perspectives into the regulatory role of transposable elements in polyploid adaptive evolution.

Phylogenomic analysis of 476 single-copy genes demonstrated that P. quinquefolius, despite its current distribution in eastern North America, exhibits the closest phylogenetic affinity to P. ginseng. Genome collinearity analysis revealed high synteny between subgenomes A and B of P. ginseng and P. quinquefolius, with their subgenome B forming a monophyletic clade with P. japonicus, P. notoginseng, and P. vietnamensis, while subgenome A constituted an independent lineage. The concordance between divergence time and syntenic patterns indicates that hybridization-coupled polyploidization directly shaped gene family expansion and evolutionary trajectories in these tetraploids. Furthermore, Ks distribution and collinearity analyses identified three critical WGD events: the core eudicot γ whole-genome triplication (WGT), followed by Panax-specific Pg-β and Pg-α WGD events. Notably, both Pg-β WGD and γ WGT events were shared between subgenomes A and B. Comparative evolutionary analysis indicates that P. quinquefolius and P. ginseng originated through recent allopolyploidization from a shared hybridization event, with subsequent geographic isolation driving species divergence in the Panax genus.

Pan-genome analysis revealed conserved and dynamic genomic architectures in P. ginseng, revealing a remarkably high core gene proportion (54.1%) that significantly exceeds those of maize and rice, suggesting strong evolutionary conservation. Wild-simulated and cultivated ginseng share a highly conserved core genomic framework, suggesting that cultivation modes and growth environments may influence medicinal quality traits through divergent regulatory mechanisms. Concurrently, heterogeneous distributions of SVs were observed among different samples. These findings provide genomic evidence for understanding P. ginseng population diversity while indicating that the conservation of core genes may limit adaptive evolution, whereas localized genetic variations could influence medicinal quality. It should be acknowledged that the current pan-genome, while representing 7 representative samples covering major production regions in China (Jilin/Liaoning) and Korean germplasm, lacks accessions from Russia and Japan. This may constrain comprehensive characterization of the species-wide genetic diversity. Importantly, as China and Korea constitute the core regions for cultivated ginseng, their germplasms sufficiently capture the primary genetic framework of domesticated populations. The highly conserved core gene set identified here provides definitive guidance for deciphering regulatory networks of key agronomic traits. Expanding to 76 representative samples through whole-genome resequencing, we identified significant associations between genetic heterogeneity and phytochemical stability across geographical populations, with the KD group exhibiting low genetic diversity but high chemical uniformity, in contrast to the FSWL population, which showed the opposite pattern. This correlation reflects the association between genetic background and medicinal quality, while conventional ISSR markers were inherently unable to resolve these associations consistently due to PCR biases and low genomic coverage. To further dissect the genetic basis of key pharmacological components in ginseng, we performed a genome-wide association study (GWAS) targeting six ginsenosides (ginsenoside Rg₁, Re, Rb₁, Rc, Ro, Rd) stipulated in the Chinese Pharmacopoeia. This analysis identified a significant SNP locus on Chr12 associated with ginsenoside Rg1 (P = 1.9187E⁻⁶, Fig. S6). Through integrated analysis of genome-wide SNPs and HPLC fingerprints, we established a genetic marker model capable of distinguishing breeding lines, registered cultivars, and wild-simulated ginseng, while developing a novel evaluation system connecting genetic diversity with chemical stability. These results not only confirm the theoretical premise that genetic diversity governs medicinal quality but also deliver practical molecular tools for cultivar authentication and lay the scientific foundation for advancing herbal medicine quality control from empirical judgment to molecular design.

In summary, this study assembled a chromosome-level genome of ginseng under wild-simulated conditions, systematically elucidating the differentiation patterns of its allotetraploid origin and the evolutionary dynamics of subgenomic repetitive sequences. Pan-genome analysis revealed the high conservation of core gene clusters in ginseng, while the integration of population genetic analysis and chemical fingerprinting established a novel evaluation system for ginseng germplasm resources. These findings not only provide a molecular basis for cultivar identification and quality control but also offer critical theoretical insights for medicinal plant genomics research and the optimization of molecular breeding strategies in ginseng.

Conclusion

To elucidate the genetic diversity of P. ginseng germplasm and evaluate its quality stability, this study systematically analyzed the genome architecture, evolutionary history, variation distribution, genetic polymorphism, and chemical consistency of ginseng. Subgenome differentiation in ginseng originated from distinct evolutionary events. Comparative genomic analysis provided critical insights into the expansion of terpenoid biosynthesis gene families and polyploidization-driven evolution. Pan-genome analysis revealed remarkable genetic conservation across ginseng accessions. Population genetic analysis identified 105 specific SNPs for reliable cultivar discrimination. The established genetic-chemical evaluation system effectively assesses ginseng quality stability, providing a scientific basis for quality standardization in medicinal ginseng cultivation and processing.

Ethics statement

This article does not contain any studies with human or animal subjects.

Compliance with ethics requirements

This article does not contain any studies with human or animal subjects.

Data availability statement

The genome assemblies and gene annotations have been deposited in the Global Pharmacopoeia Genome Database (GPGD) [72] at: https://www.gpgenome.com/species/606. Additionally, supporting datasets are contained within the Supplementary materials.

CRediT authorship contribution statement

Xuejiao Liao: Original draft, Validation, Investigation, Formal analysis, Project administration Yidan Xi: Investigation, Validation, Data curation. Baosheng Liao: Methodology, Resources, Data curation. Shuiming Xiao: Methodology, Data curation. Shuai Guo: Methodology, Investigation. Haoyu Hu: Software, Visualization, Formal analysis. Zhihai Huang: Supervision, Editing. Deqiang Dou: Conceptualization, Funding acquisition, Supervision. Shilin Chen: Formal analysis, Project administration, Supervision. Jiang Xu: Conceptualization, Funding acquisition, Project administration, Resources.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: