search
Contact Us
HOME

  • Crossref logo
  • Crossref Similarity Check logo
Journal Search Engine

to

Search Advanced Search Adode Reader(link)
Download PDF Export Citaion korean bibliography PMC previewer
ISSN : 2287-5824(Print)
ISSN : 2287-5832(Online)
Journal of The Korean Society of Grassland and Forage Science Vol.46 No.3 pp.255-263
DOI : https://doi.org/10.5333/KGFS.2026.46.3.255

Development of an NGS-Derived SNP Barcode for Identification of the Forage Barnyard Millet (Echinochloa esculenta) variety ‘Jo-on’

Bo Ram Choi, Ki-Won Lee, Yowook Song, Chang-Woo Min*
Forages Production Systems Division, National Institute of Animal Science, RDA, Cheonan 31000, Republic of Korea
*Corresponding author: Chang-Woo Min, Forages Production Systems Division, National Institute of Animal Science, RDA, Cheonan 31000, Republic of Korea. Tel: +82-41-580-6755, Fax: +82-41-580-6779, E-mail: mcw1004@korea.kr
September 23, 2026 September 28, 2026 September 29, 2026

Abstract


Barnyard millet (Echinochloa esculenta) is an important summer forage crop in Korea, but reliable molecular tools for variety identification remain limited. This study aimed to establish genome-wide single-nucleotide polymorphism (SNP) resources and select barcode candidates for discriminating the forage variety ‘Jo-on’ from ‘Borajik’ and ‘Jejupi’. Whole-genome resequencing of nine barnyard millet genotypes yielded 523,495 high-confidence homozygous SNPs after stringent filtering. Genome-wide variation resolved four genetic groups and showed that ‘Jo-on’ was clearly differentiated from ‘Borajik’ and ‘Jejupi’. Comparison of allele combinations among the three target varieties identified two complementary classes of SNPs. When one locus from each class was combined, ‘Jo-on’, ‘Borajik’, and ‘Jejupi’ were assigned distinct two-locus profiles of bb, ab, and aa, respectively. Thirty representative loci were selected from different polymorphism classes and contig groups to provide alternatives for downstream assay design. These results demonstrate that a small combination of informative SNPs can discriminate the three target varieties despite the close genetic relationships within barnyard millet germplasm. The selected loci should therefore be regarded as barcode candidates rather than validated diagnostic markers. Validation of primer specificity, genotype reproducibility, and within-variety uniformity across independent plants and seed lots will be required before application to variety authentication and seed purity assessment.


Barnyard millet , Variety identification , Jo-on , NGS , SNP barcode

초록

    Ⅰ. INTRODUCTION

    Barnyard millet (Echinochloa esculenta) is a summer annual C4 forage crop that grows well under high-temperature and poorly drained soil conditions (Yabuno, 1987). Its short growing period of 60–80 days and tolerance to wet conditions make it suitable for paddy-field cultivation and double cropping with winter forage crops, thereby helping to alleviate seasonal gaps in forage production (Lim et al., 2021;Park et al., 2022).

    In Korea, the varieties of E. esculenta protected by the Korea Seed and Variety Service have primarily been developed for food or processing purposes. To expand the use of barnyard millet as a summer forage crop, the National Institute of Animal Science developed the early-maturing variety ‘Jo-on’ for cultivation in paddy fields. As ‘Jo-on’ enters seed multiplication and distribution, a reliable identification method is required to protect the variety, maintain the genetic purity of commercial seed lots, and prevent admixture with other varieties. However, agronomic traits such as plant type, plant height, and yield can vary depending on the cultivation environment and growth stage (Lee et al., 2013). Molecular markers are therefore needed to complement phenotype-based distinctness, uniformity, and stability evaluation.

    Previous DNA marker studies of Echinochloa have focused mainly on species discrimination and germplasm diversity. Simple sequence repeat (SSR) markers have been used to distinguish E. esculenta from its wild relative, E. crus-galli, and to assess genetic diversity among accessions and breeding lines (Nozawa et al., 2006;Meniya et al., 2023;Joshi et al., 2026). To the best of our knowledge, however, molecular markers developed specifically for identifying registered E. esculenta varieties have not been reported. Moreover, E. esculenta is an allohexaploid species ( 2 n = 6 x = 54 ) comprising three subgenomes (Satomura, 2025), which may complicate SSR genotyping because of cross-amplification among homoeologous loci and difficulties in interpreting allele dosage. Single-nucleotide polymorphism (SNP) markers offer an alternative because they are abundant, locus-specific, highly reproducible, and compatible with high-throughput genotyping platforms.

    Therefore, this study aimed to develop a next-generation sequencing (NGS)-based SNP barcode for the identification of the Korean forage barnyard millet variety ‘Jo-on’. The candidate markers are expected to support variety protection and genetic purity management during seed multiplication and distribution.

    Ⅱ. MATERIALS AND METHODS

    1. Plant materials

    This study evaluated nine barnyard millet genotypes comprising varieties, locally maintained materials, and breeding lines developed by the National Institute of Animal Science, Rural Development Administration, Republic of Korea. The plant materials included ‘WS2 (Shirohie)’, ‘WS28 (Mochihie)’, ‘WS30 (Taiwan)’, ‘WS31 (Jeju_ecotype2)’, ‘F2_20BMCB1’, ‘F4_20BMCB2_R-R’, ‘F4_20BMCB2_G-G(Jo-on)’, ‘Borajik’, ‘Jejupi’. All plant materials were grown for three weeks, after which young leaves were collected from each plant. To preserve DNA integrity, the collected leaves were immediately flash-frozen in liquid nitrogen. Frozen leaf samples were stored at -70°C in an ultra-low temperature freezer until genomic DNA extraction.

    2. DNA extraction and sequencing data production

    Genomic DNA was extracted from young leaf tissues of each of the nine barnyard millet varieties using the CTAB method (Kidwell and Osborn, 1992). To verify DNA quality, 3 µL of extracted DNA was subjected to electrophoresis at 200 V for 25 minutes, and DNA integrity was confirmed using a 1 kb+ ladder. For whole-genome-sequencing (WGS) library preparation, 100 µg of gDNA from each of the nine barnyard millet varieties was fragmented, treated at 32°C for 15 minutes, and then fragmentation enzymes were inactivated at 65°C for 30 minutes. Adapter ligation was performed at 20°C for 15 minutes, followed by purification with a QIAquick PCR Purification Kit (QIAGEN Inc, Valencia, CA, USA). PCR was conducted for 8 cycles using a GeneExplorer (BIOER, Hangzhou, China), and the band quality was assessed using a 100 bp ladder before NGS library construction.

    Finally, sequencing of the prepared libraries was performed at 2 × 150 bp using the Illumina HiSeq X platform.

    3. Sequencing data preprocessing and SNP analysis

    Raw WGS reads were preprocessed using SolexaQA (version 1.13) and Trimmomatic (version 0.39) (Cox et al., 2010;Bolger et al., 2014). Reads with a Phred score below 20 and reads shorter than 25 bp were removed using DynamicTrim and LengthSort in SolexaQA, and residual bases with Phred quality scores below 20 were trimmed using Trimmomatic. The cleaned reads were mapped to the Echinochloa crus-galli reference genome using BWA-MEM (version 0.7.17) (Li, 2013). The reference assembly was ASM90020540v1 (GCA_900205405.1), comprising 4,534 scaffolds and totaling approximately 1.49 Gbp according to the NCBI assembly record.

    SNPs were called against the reference genome using SAMtools (version 0.1.16), and varFilter was applied with a minimum mapping quality (−Q) of 30, minimum read depth (−d) of 3, and maximum read depth (−D) of 611 (Li et al., 2009). Variants adjacent to other SNPs or insertions/deletions (InDels) were excluded. Genotypes with a reference or alternative allele proportion of ≥90% were classified as homozygous, whereas those with a non-reference allele proportion of >10% and <90% were classified as heterozygous. SNP sites from the nine WGS accessions were merged by reference position to construct the genotype matrix. Loci were retained when the missing data rate was 0%, MAF was >20%, and the heterozygous genotype rate was 0%.

    SNP loci were annotated using the genome feature annotation tracks associated with the Echinochloa crus-galli reference. Each SNP was represented as “a” for the reference allele, “b” for the alternative allele, “h” for a heterozygous genotype, and “−” for missing data. Candidate loci were retained where homozygous genotypes were observed in ‘F4_20BMCB2_G-G(Jo-on)’, whereas polymorphism was observed among the other varieties. From these candidates, the minimum SNP combination required to distinguish ‘F4_20BMCB2_G-G(Jo-on)’ from ‘Borajik’ and ‘Jejupi’ was determined, and the corresponding loci were designated as variety-discriminating marker candidates. Candidates were classified into three MAF/PIC classes. Contigs were then grouped operationally according to the first nine characters of their accession identifiers (OAMR01000–OAMR01004). Within each barcode number, MAF/PIC class, and contig group, loci were sorted by full contig accession and POS, and the first locus was selected, resulting in 30 representative SNP candidates.

    4. Phylogenetic tree and principal component analysis (PCA)

    Genetic relationships among the nine barnyard millet genotypes were evaluated using all 523,495 filtered SNP loci. Pairwise SNP p-distance was calculated as the proportion of loci with discordant homozygous genotypes, and a neighbor-joining tree was constructed from the resulting distance matrix (Saitou and Nei, 1987). Branch support was assessed with 1,000 bootstrap replicates (Felsenstein, 1985). To reduce inflation caused by linked SNPs within the same contig, the 1,704 contigs containing at least one of the 523,495 filtered SNP loci among the 4,534 reference scaffolds were sampled with replacement while retaining all loci in each selected contig.

    For PCA, homozygous reference and alternative genotypes were encoded as 0 and 1, respectively, and each SNP locus was centered by its mean genotype. Principal components were obtained by eigen decomposition of the sample covariance matrix. K-means partitions from K=2 to K=8 were evaluated using Euclidean distances in the complete eight-dimensional principal-component space, and the mean silhouette coefficient was calculated for each K (Rousseeuw, 1987). The main cluster solution was selected by considering the silhouette coefficient, the occurrence of singleton clusters, and concordance with the PCA and neighbor-joining patterns. Analyses were performed with a custom Python 3.12.14 script using NumPy 2.3.5.

    Ⅲ. RESULTS AND DISCUSSION

    1. Sequencing data production and preprocessing

    WGS of nine barnyard millet genotypes generated 28.2 Gb for ‘F4_20BMCB2_R-R’ to 39.1 Gb for ‘Jo-on’ (Table 1). Based on the 1.49 Gb reference genome, these sequence outputs corresponded to approximately 18.9–26.2-fold genome coverage. After quality filtering and preprocessing, approximately 67.70–90.54% of the reads were retained. The amount of clean sequence data ranged from 22.0 Gb for ‘F2_20BMCB1’ to 27.2 Gb for ‘WS28 (Mochihie)’, while the mapping rates ranged from 92.92 to 99.77%. These results indicated that the sequence data provided sufficient coverage and mapping quality for subsequent SNP detection.

    2. SNP identification, filtering and genetic relationships

    The WGS reads from each barnyard millet sample were aligned to the reference genome, and the raw SNP positions identified across all samples were combined to construct a union set of candidate loci. At loci where no SNP was detected in a sample, the corresponding reference allele was assigned, whereas low-confidence or incorrectly called loci were excluded.

    An average of 4,817,510 SNP loci per variety was identified, comprising 2,111,424 homozygous alternative SNPs and 2,706,086 heterozygous SNPs (Table 2). Across all nine genotypes, heterozygous SNP calls accounted for more than 50% of the total SNP calls, which may partly reflect the contribution of homoeologous reads to apparent heterozygous calls when subgenome-aware read categorization is not applied in an allohexaploid genome (Page and Udall, 2015). ‘Jejupi’ showed a higher heterozygous SNP proportion (68.54%) than the other genotypes (53.86–55.67%), possibly reflecting residual genetic heterogeneity in this locally maintained landrace without a documented history of pure-line selection, although this interpretation requires validation across multiple individuals. Sequential filtering of 8,954,698 union SNP sites retained 7,830,927 loci without missing genotypes, 1,436,555 loci with MAF >20%, 1,184,573 loci satisfying both criteria, and finally 523,495 loci without heterozygous calls (Table 3). These 523,495 loci were used for all subsequent PCA, clustering, and genetic-distance analyses.

    The mean silhouette coefficient based on the complete PCA space was highest at K=5 (0.66950), but exceeded that at K=4 (0.66435) by only 0.00515 (Fig. 1). Although the marginally higher value at K=5 indicated a slight improvement in the balance between within-cluster cohesion and between-cluster separation, this solution separated ‘WS31 (Jeju_ecotype2)’ from ‘Borajik’ and ‘Jejupi’ as a singleton cluster. In silhouette analysis, singleton observations are conventionally assigned a coefficient of zero, and a high overall coefficient should be interpreted in conjunction with the resulting cluster configuration (Rousseeuw, 1987). Therefore, the small increase observed at K=5 was not considered sufficient evidence for five stable genetic groupings. K=4 produced no singleton clusters and showed a higher silhouette coefficient in the PC1–PC2 plane than K=5, with values of 0.87854 and 0.80347, respectively. Accordingly, K=4 was adopted as the more parsimonious working solution. At K=4, ‘WS2 (Shirohie)’ clustered with ‘Jo-on’, ‘WS28 (Mochihie)’ with ‘F2_20BMCB1’, ‘WS30 (Taiwan)’ with ‘F4_20BMCB2_R-R’, and ‘WS31 (Jeju_ecotype2)’ with ‘Borajik’ and ‘Jejupi’. PC1 and PC2 explained 39.35 and 27.47% of the total variation, respectively, accounting for 66.82% of the cumulative variation (Fig. 2).

    Among the nine genotypes, ‘Borajik’ and ‘Jejupi’ were selected for direct comparison with ‘Jo-on’ because they are commonly distributed barnyard millet materials in the domestic seed market. Pairwise SNP p-distances among the three focal varieties were 0.43416 between ‘Jo-on’ and ‘Borajik’, 0.43296 between ‘Jo-on’ and ‘Jejupi’, and 0.34967 between ‘Borajik’ and ‘Jejupi’. Thus, ‘Borajik’ and ‘Jejupi’ were genetically closer to each other, whereas ‘Jo-on’ showed a similar degree of differentiation from both varieties. The neighbor-joining topology also separated ‘Jo-on’ from ‘Borajik’ and ‘Jejupi’, which were positioned on distinct branches (Fig. 3). Collectively, the PCA, pairwise-distance, and neighbor-joining results consistently supported the discrimination of the three focal varieties within the analyzed panel.

    3. Extraction of marker candidates for variety discrimination

    SNP genotypes were compared among ‘Jo-on’, ‘Borajik’ and ‘Jejupi’ in that order, to identify complementary barcode candidates. Barcode number 1 comprised 95,179 loci with a ‘baa’ pattern, whereas barcode number 2 comprised 64,366 loci with a ‘bba’ pattern. Accordingly, combining one locus from each barcode set generated distinct two-locus profiles of ‘bb’, ‘ab’, and ‘aa’ for ‘Jo-on’, ‘Borajik’, and ‘Jejupi’, respectively (Table 4).

    All candidate loci were homozygous and complete across the nine genotypes. In barcode number 1, 59,400, 9,722, and 26,057 loci belonged to the MAF/PIC classes (22.2%, 0.286), (33.3%, 0.346), and (44.4%, 0.372), respectively; the mean MAF and PIC values were 29.4% and 0.316. The corresponding counts for barcode number 2 were 2,846, 47,140, and 14,380, with mean MAF and PIC values of 35.3% and 0.349. From each barcode number and MAF/PIC class, one representative locus was selected from each of the five contig groups, resulting in 30 representative SNP candidates (Table 5).

    For biallelic SNPs, PIC reaches a theoretical maximum of 0.375 at balanced allele frequencies under the equation proposed by Botstein et al. (1980). The mean PIC values therefore indicate that both candidate sets were informative within the nine accession panel. However, MAF and PIC represent panel-level informativeness and do not demonstrate assay conversion, genotype repeatability, or fixed specificity for ‘Jo-on’. The minimal SNP panel developed for apple was evaluated based on low failure and error rates and concordant profiles among replicate samples (Winfield et al., 2020). Similarly, the cucumber core SNP set was reduced from genome-wide candidates after considering conserved flanking sequences and experimental genotyping performance (Zhang et al., 2020). Therefore, the 30 loci presented in Table 5 should be regarded as representative candidates. Final marker selection should consider MAF, PIC, flanking-sequence uniqueness, primer design, genotype repeatability, and validation across multiple individuals, seed lots, and additional closely related varieties.

    Ⅳ. CONCLUSION

    Genome-wide analysis of 523,495 filtered SNP loci showed that ‘Jo-on’ was distinguishable from ‘Borajik’ and ‘Jejupi’ within the nine genotype panel. Two complementary candidate sets were identified. The set designated as barcode number 1 showed a ‘baa’ pattern, while barcode number 2 showed a ‘bba’ pattern in the genotype order ‘Jo-on’–‘Borajik’–‘Jejupi’. Combining one locus from each set assigned unique ‘bb’, ‘ab’, and ‘aa’ profiles to ‘Jo-on’, ‘Borajik’, and ‘Jejupi’, respectively. Thirty representative candidates were selected across three MAF and PIC classes and five contig groups. These results provide a basis for constructing a minimal SNP barcode for distinguishing ‘Jo-on’ from ‘Borajik’ and ‘Jejupi’. However, independent assay conversion and validation across multiple individuals, seed lots, and closely related varieties are required before these candidates can be established as variety-identification markers.

    Ⅴ. ACKNOWLEDGEMENT

    This study was supported by the Cooperative Research Program for Agriculture Science & Technology Development (Project No. PJ01669905) and the 2026 RDA Fellowship Program of the National Institute of Animal Science funded by RDA, Republic of Korea.

    Figure

    KGFS-46-3-255_F1.jpg

    Mean silhouette coefficients for K=2 to 8 based on filtered SNP loci.

    KGFS-46-3-255_F2.jpg

    Principal component analysis of nine barnyard millet genotypes based on filtered SNP loci.

    KGFS-46-3-255_F3.jpg

    Neighbor-joining tree of nine barnyard millet genotypes based on pairwise SNP p-distances from filtered SNP loci.

    Table

    Summary statistics of whole-genome sequencing data after preprocessing for nine barnyard millet genotypes
    1Clean reads/Raw reads (%): (Total number of clean reads / Total number of raw reads) × 100
    2Mapping rate(%): Number of mapped reads/Number of clean reads × 100
    Summary of SNP counts by zygosity in nine barnyard millet genotypes
    ¹Average: mean SNP count across the 9 barnyard millet genotypes
    ²Total: sum of SNP count across the 9 barnyard millet genotypes
    Sequential filtering of the SNP matrix across nine barnyard millet genotypes
    ¹Missing rate = 0% : only loci without missing genotype calls across all accessions were retained.
    ²MAF (minor allele frequency) > 20% : SNP loci with a MAF greater than 20% across all samples were retained.
    ³Heterozygous = 0% : only loci without heterozygous genotype calls across all accessions were retained.
    Barcode-SNP candidate sets for discrimination among ‘Jo-on’, ‘Borajik’, and ‘Jejupi’
    a, same nucleotide as the reference genome; b, different nucleotide from the reference genome.
    Nucleotide information, minor allele frequency (MAF), and polymorphism information content (PIC) of representative barcode-SNP candidates for identification of three barnyard millet varieties
    *Contigs were grouped according to the first nine characters of their accession identifiers (OAMR01000–OAMR01004).
    **POS, position; REF, reference allele; ALT, alternative allele.

    Reference

    1. Bolger, A.M., Lohse, M. and Usadel, B. 2014. Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics. 30(15):2114-2120.
    2. Botstein, D., White, R.L., Skolnick, M. and Davis, R.W. 1980. Construction of a genetic linkage map in man using restriction fragment length polymorphisms. The American Journal of Human Genetics. 32(3):314-331.
    3. Cox, M.P., Peterson, D.A. and Biggs, P.J. 2010. SolexaQA: At-a-glance quality assessment of Illumina second-generation sequencing data. BMC Bioinformatics. 11:485.
    4. Felsenstein, J. 1985. Confidence limits on phylogenies: An approach using the bootstrap. Evolution. 39(4):783-791.
    5. Joshi, D.C., Kumar, S., Hasan, N., Rani, V., Bhinda, M.S. and Kant, L. 2026. Development and validation of genome-wide simple sequence repeat markers for large scale genotyping applications in barnyard millet. BMC Plant Biology.
    6. Kidwell, K.K. and Osborn, T.C. 1992. Simple plant DNA isolation procedures. In: J.S. Beckmann and T.C. Osborn (Eds.), Plant genomes: Methods for genetic and physical mapping. Springer. Dordrecht, Netherlands. pp. 1-13.
    7. Lee, J.J., Kim, J.G., Sung, B.R., Song, T.H. and Park, T.S. 2013. Studies on growth, forage yield, and nutritive value according to different seeding dates of barnyard millet. Journal of the Korean Society of Grassland and Forage Science. 33(4):245-251.
    8. Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, G., Durbin, R. and 1000 Genome Project Data Processing Subgroup. 2009. The sequence alignment/map format and SAMtools. Bioinformatics. 25(16):2078-2079.
    9. Li, H. 2013. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arXiv. arXiv:1303.3997.
    10. Lim, E.A., Lee, K.W., Choi, B.R., Park, H.S., Woo, J.H., Kim, D.I. and Lee, S.H. 2021. A study on features of forage barnyard millet and related research trends. Journal of the Korean Society of Grassland and Forage Science. 41(3):217-222.
    11. Meniya, V.H., Padhiyar, S.M., Desai, H., Kheni, J.K. and Tomar, R.S. 2023. Development and validation of microsatellite markers for barnyard millet obtained by partial genome assembly. Annals of Arid Zone. 62:83-89.
    12. Nozawa, S., Takahashi, M., Nakai, H. and Sato, Y. 2006. Difference in SSR variations between Japanese barnyard millet and its wild relative E. crus-galli. Breeding Science. 56:335-340.
    13. Page, J.T. and Udall, J.A. 2015. Methods for mapping and categorization of DNA sequence reads from allopolyploid organisms. BMC Genetics. 16(Suppl. 2):S4.
    14. Park, H.S., Choi, K.C., Yang, S.H., Jung, J.S. and Lee, B.H. 2022. Evaluation of growth characteristics and yield potential of summer emergency forage crops. Journal of the Korean Society of Grassland and Forage Science. 42(1):26-31.
    15. Rousseeuw, P.J. 1987. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics. 20:53-65.
    16. Saitou, N. and Nei, M. 1987. The neighbor-joining method: A new method for reconstructing phylogenetic trees. Molecular Biology and Evolution. 4:406-425.
    17. Satomura, K. 2025. Tribe Paniceae cereals with different ploidy levels. Genes. 16:426.
    18. Winfield, M., Burridge, A., Ordidge, M., Harper, H., Wilkinson, P., Thorogood, D., Copas, L., Edwards, K. and Barker, G. 2020. Development of a minimal KASP marker panel for distinguishing genotypes in apple collections. PLoS ONE. 15(11):e0242940.
    19. Yabuno, T. 1987. Japanese barnyard millet (Echinochloa utilis, Poaceae) in Japan. Economic Botany. 41:484-493.
    20. Zhang, J., Yang, J., Zhang, L., Luo, J., Zhao, H., Zhang, J. and Wen, C. 2020. A new SNP genotyping technology Target SNP-seq and its application in genetic analysis of cucumber varieties. Scientific Reports. 10:5623.
    Copyright © The Korean Society of Grassland and Forage Science. All rights reserved.