Dactylis glomerata L., commonly known as Orchardgrass is a member of the gramineous family Poaceae (Mao et al., 2016; Zhang et al., 2022). The genus Dactylis is distinctive and distinctly different from other genera in the Poaceae family, demonstrating significant variance in taxonomic characteristics and flourishing in a wide range of environments (Lumaret, 1997). Despite having just one species, D. glomerata, the genus Dactylis has at least eighteen subspecies (Sanada et al., 2010). It is regarded as the fourth most significant wild forage grass globally (Chtourou-Ghorbel et al., 2024). This perennial cool-season grass grows natively in North Africa, West and Central Asia, the Mediterranean basin, and Europe (Stewart and Ellison, 2010; Xie et al., 2014; Lee et al., 2021).

Diploid, tetraploid, and some hexaploid populations of D. glomerata occur naturally (Lolicato and Rumball, 1994; Sanada et al., 2010). Most diploid populations have limited distributions and are found in specific regions, collectively representing about 5% of all wild Dactylis. In contrast, tetraploids are widespread, occurring continuously throughout temperate regions of Europe, the Middle East, West and Central Asia, and North Africa (Lumaret and Borrill, 1988). Both ploidy types often coexist in certain areas while hexaploids are confined to restricted areas in Libya and Western Egypt (Jones and Borrill, 1962; Last et al., 2013). The species exhibits an infraspecific polyploid complex in grasses, resulting from autopolyploidy due to polysomic inheritance. Polyploid populations possess evolutionary advantages stemming from greater heterozygosity and reduced inbreeding depression, facilitating enhanced colonization and adaption to fluctuating ecological conditions (Soltis and Soltis, 2000; Van de Peer et al., 2021).

This is an important fodder crop due to its favorable palatability and high sugar content for animals, supporting animal feed and enhancing dairy and meat production in temperate climates (Wilkins and Humphreys, 2003; Katoch, 2022). It can withstand shade, barrenness, and drought. Furthermore, the rapid growth of its root system makes it particularly valuable as a cover crop for preventing surface erosion and rehabilitating degraded soils (Costa et al., 2016; Copăcean et al., 2019). Likewise, its deep root system allows it to access water and nutrients, making it drought-tolerant and providing high-quality forage for efficient production (Gaier et al., 2024).

Genetic diversity is a critical resource in any breeding program (Iqbal et al., 2023; Altaf et al., 2024a; Yalinkiliç et al., 2024) aimed at improving traits such as yield, disease resistance, abiotic stress tolerance, and forage quality (Jovovic et al., 2020; Baloch et al., 2024). Genetic variations within populations enable breeders to select desirable traits and develop varieties suited to specific environmental conditions or management practices (Bunjkar et al., 2024; Altaf et al., 2024b). Given the widespread cultivation of D. glomerata, understanding the genetic heterogeneity within germplasm collections is essential for breeding programs focused on enhancing adaptability, resilience, and productivity (Zhang et al., 2022). Despite its significance, D. glomerata faces substantial challenges from both biotic and abiotic stresses, which are likely to be exacerbated by climate change. This underscores the urgency of developing resilient cultivars to ensure its continued productivity and adaptability (Shahzad et al., 2021). Breeding for disease resistance is particularly important to maintain forage yield and quality (Capstaff and Miller, 2018). Traditional phenotypic selection is slow and resource-intensive, whereas modern marker-assisted selection (MAS) provides targeted approaches that rely on understanding genetic diversity and key traits (Nadeem et al., 2018). Different types of molecular markers (SSR, simple sequence repeat; AFLP, amplified fragment length polymorphism; DArT, diversity arrays technology; ISSR, inter simple sequence repeats; and RAPD, randomly amplified polymorphic DNA) have been developed and successfully applied in the marker-assisted breeding program of various crops (Nadeem et al., 2018). However, molecular markers that are resolved through gel-based electrophoresis face limitations such as low reliability, limited genome coverage, labor-intensive processes, and high costs or sequence needs (Jaccoud et al., 2001). These drawbacks affect their use in many crops, particularly ‘orphan’ crops and polyploid species (Sánchez-Sevilla et al., 2015).

New markers developed through Next Generation Sequencing platforms (NGS) are the current prime alternative for molecular studies since they cover a vast proportion of the genome (Kilian et al., 2012). Genotyping-by-sequencing (GBS) is a prevalent NGS technique designed for the concurrent identification of novel markers and genotyping of specific germplasm (Elshire et al., 2011; Rayaprolu et al., 2022). This technology is high-throughput and cost-effective, utilized across multiple crops for diverse applications (Elshire et al., 2011; Poland and Rife, 2012; Huang et al., 2014; Alipour et al., 2017; Geleta et al., 2020; Baloch and Nadeem, 2022). The predominant DNA markers produced with the GBS approach are single nucleotide polymorphisms (SNPs). These markers signify the predominant sequence-based variations within crop genomes, rendering them exceptionally appropriate for the examination of genetic variability, marker-trait association, population structure, mapping quantitative trait loci (QTL), genomic selection, map-based cloning, and various plant breeding applications (Batley and Edwards, 2007; Hiremath et al., 2012; Tsehay et al., 2020; Gardoce et al., 2023; Altaf et al., 2024c).

Advancements in sequencing technologies have significantly advanced statistical genetic methods, with genome-wide association studies (GWAS), enabling the identification of genes or alleles linked to specific traits (Sahito et al., 2024; Ahmed et al., 2024). GWAS utilizes next-generation sequencing (NGS) data, to examine thousands of genetic variants, most commonly single nucleotide polymorphisms (SNPs), across diverse genomes to identify those statistically associated with specific traits. This methodology surpasses the limitations of traditional gene and quantitative trait loci (QTL) mapping, which rely on biparental crosses and are constrained by limited allelic diversity and genomic resolution (Borevitz and Nordborg, 2003). Unlike traditional methods, GWAS utilizes phenotypically characterized germplasm collections, eliminating the need for structured populations (e.g., F2 or F3 progenies), while providing greater allelic diversity and higher mapping resolution, sometimes to the gene level (Tibbs Cortes et al., 2021). This approach has been instrumental in identifying QTL that explains substantial phenotypic variation in a variety of plant traits. Once a phenotype-marker association is identified, downstream analyses of candidate genes in nearby genomic regions can yield valuable insights into the underlying biology of the trait. It has been successfully employed in various plants, including Arabidopsis, rice, soybean, wheat, cotton, sorghum, common bean, and Orchardgrass, to study traits ranging from yield to abiotic and biotic stress tolerance (Liu and Yan, 2019; Sukumaran and Yu, 2014; Varshney et al., 2012; Xu et al., 2024; Altaf et al., 2024d).

Studies assessing genetic diversity in D. glomerata using SSR, restriction fragment length polymorphism (RFLP), and start codon targeted (SCoT) markers are documented (Madesis et al., 2014; Jiang et al., 2015; Costa et al., 2016), but the use of SNP markers for this purpose remains limited. No studies have yet utilized the GBS approach for SNP-based genetic diversity analysis in this crop. The present study aimed to evaluate the genetic diversity and population structure of 91 accessions of D. glomerata using GBS-based SNP markers.

The GBS analysis yielded, after data filtering, a total of 2, 913 high-quality SNPs across 91 D. glomerata accessions. The genetic diversity indices for each geographical origin were measured based on provinces ( Table 1 ). There was a noticeable range in mean genetic diversity (GD), polymorphism information index (PIC), observed heterozygosity (Ho), and expected heterozygosity (He) values across provinces. Accessions from Erzurum exhibited the highest GD, with a mean of 0.26 and He of 0.5328, and PIC value (0.21), indicating significant genetic variability within this province. Similarly, accessions from Çanakkale and Sivas also showed high diversity levels ( Supplementary Table S2 ).

The analysis of molecular variance (AMOVA) on provincial level revealed that the majority of the genetic variation (95.4%) was attributed to differences among populations based on provincial divisions. In contrast, a smaller proportion of the variation (4.6%) occurred among individual accessions within populations ( Table 2 ). The overall pair-wise genetic differentiation (FST) was 0.143 ( Table 2 ).

SNP markers data was used for genetic structure analysis, using the Bayesian clustering model implemented in the STRUCTURE software. Selection of the optimum number of populations (K) was done using the post hoc methods, based on their mean likelihood and their delta K ( Figure 2 ). The structure divided the studied germplasm into 3 populations ( Figure 3A ). The pair-wise genetic dissimilarity values among 91 D. glomerata accessions, based on 2,937 SNP markers, were calculated using Hamming distance to generate a genetic distance matrix. This matrix was used to construct a dendrogram with the UPGMA. The UPGMA analysis divided the studied accessions into three main clusters (I, II, and III), with further sub-structuring observed within clusters I and II, as indicated by bold black and gray dashed lines in Figure 3B . Cluster I comprised 32 accessions (34.04%), which were further divided into two sub-clusters: Ib and Ia. Sub-cluster Ib contained 20 accessions, including those from various provinces such as Antalya, Kayseri, Muğla, and Erzurum. Specifically, these accessions were Türkiye1, Antalya1, Kayseri1, Gümüşhane1, Samsun1, Türkiye2, Muğla3, Antalya2, Türkiye4, Kars8, Türkiye12, Erzurum6, Türkiye19, Tokat1, İstanbul2, Türkiye31, Türkiye32, Türkiye33, Malatya1, and Tahran1. Sub-cluster Ia consisted of 10 accessions from Kars, Tunceli, Ankara, Sivas, Erzurum, and Ağrı, including Kars1, Kars3, Tunceli1, Ankara1, Kayseri2, Sivas2, Ankara2, Erzurum5, Türkiye7, and Ankara3.

Cluster II contained a total of 22 accessions (23.40%), which were also divided into two sub-clusters, IIb and IIa. Sub-cluster IIb consisted of 5 accessions, including Erzurum2, Türkiye3, İstanbul1, Türkiye16, and Türkiye29. The remaining 17 accessions formed sub-cluster IIa, encompassing accessions from Sivas, Bursa, Aydın, Muğla, and other provinces, such as Sivas3, Bursa1, Aydın1, Muğla1, Türkiye9, Türkiye11, Türkiye14, Türkiye18, Türkiye20, Türkiye21, Türkiye22, Türkiye25, Çanakkale1, Çanakkale2, Türkiye27, Türkiye28, and Ayfam ( Figure 3B ).

Cluster III was the largest, containing 40 accessions (42.55%). This cluster included accessions from diverse locations such as Erzurum, Ardahan, Kars, Yozgat, and Bingöl, with notable entries being Erzurum1, Ardahan1, Kars2, Ardahan2, Afyon1, Kars4, Kars5, Yozgat1, Erzincan3, Sivas1, Erzurum3, Kayseri3, Kars6, Erzurum4, Muğla2, Gümüşhane2, Kars7, Türkiye5, Türkiye6, Türkiye8, Ardahan3, Kars10, Kars11, Türkiye15, Türkiye10, Türkiye13, Türkiye23, Türkiye24, Bayburt1, Türkiye26, Türkiye17, Bingöl1, Bingöl2, Bingöl3, Bingöl4, Kayseri4, Kayseri5, Elazığ1, Erdebil1, and Sivas4 ( Figure 3B ). The Principal Coordinates Analysis (PCoA) categorized the germplasm into three distinct groups, consistent with the UPGMA clustering. Accessions within these groups were visually represented using three different colors: red for Group 1, green for Group 2, and blue for Group 3, highlighting their genetic differentiation ( Figure 4 ). We collected comprehensive phenotypic data for the germplasm, with a specific focus on dry biomass weight (DBW). Through UPGMA clustering analysis, we found that accessions with similar average DBW values were closely grouped, suggesting a relationship between genetic clustering patterns and phenotypic similarity. To investigate this relationship further, we constructed histograms showing the mean DBW of all the accessions within each UPGMA-based cluster. The results indicated that Cluster II had the highest mean DBW, being significantly greater (p<0.05) than the mean DBW of Clusters I and III, whereas Cluster III exhibited the lowest mean DBW ( Figure 5 ).

The diversity indices were also calculated based on clustering and Table 3 summarizes the genetic diversity statistics for the three major clusters identified. Cluster I consisted of 28 accessions (29.8% of the total) and exhibited a mean genetic diversity (GD) of 0.22, with a polymorphic information content (PIC) of 0.19 and a minor allele frequency (MAF) of 0.14. Observed heterozygosity (Ho) was 0.23, and expected heterozygosity (He) was 0.2369. The inbreeding coefficient (Fi) was slightly negative (-0.03). Cluster II comprised 23 accessions (24.5%), and genetic diversity metrics with a mean GD of 0.22, PIC of 0.18, and MAF of 0.14. The Ho was 0.23, and He was 0.2461, slightly higher than in Cluster I. The negative inbreeding coefficient (Fi = -0.07). Cluster III was the largest cluster, including 40 accessions (42.6%). The genetic diversity metrics were slightly lower than those of the other clusters, with a mean GD of 0.21, PIC of 0.18, and MAF of 0.14. The observed and expected heterozygosity were 0.21 and 0.2121, respectively, with a near-zero inbreeding coefficient (Fi = -0.01. Furthermore, the AMOVA was also calculated on cluster base and revealed that the majority of genetic variation (85.8%) occurred within clusters, while 14.2% of the variation was attributed to differences among clusters ( Table 3 ). The overall FST value was 0.007.

The mixed linear model (MLM; Q + K) was employed to identify marker-trait associations for dry biomass weight in the D. glomerata germplasm. Notably, markers DArT-100715788 and DArT-101043591 were located to Chromosomes 1 and 6, respectively ( Figure 6 ; Table 4 ). Additionally, two markers, DArT-101171265 and DArT-101090822, both localized on Chromosome 7, demonstrated significant associations with dry biomass weight, with p-values of 5.44 × 10^-5 and 1.04 × 10^-4, respectively ( Table 4 ). These findings highlight specific genomic regions linked to dry biomass accumulation.

Genetic diversity is a critical factor in the study of any species as it directly influences the evolutionary potential and adaptability of populations (Futuyma, 2017). In the context of D. glomerata, understanding genetic diversity is essential for characterizing accessions, identifying duplications within germplasm collections, and selecting suitable parental genotypes for breeding programs. Historically, morphological traits have been used to estimate genetic similarity in species, including orchardgrass (Gauthier et al., 1999). However, phenotypic assessments are often unreliable indicators of genetic variation due to environmental influences on phenotype expression. Consequently, molecular approaches offer a more precise method for assessing genetic diversity, providing robust insights into the genetic architecture of D. glomerata, which is pivotal for its conservation and improvement.

A total of 2,913 high-quality SNPs was used to assess 91 accessions, resulting in diverse genetic indices across provinces. The highest genetic diversity was observed in accessions from Erzurum (GD = 0.26, He = 0.5328, PIC = 0.21), indicating a rich genetic pool, suggesting that environmental and geographical conditions of Erzurum may favor higher genetic variability. Conversely, lower genetic diversity (GD = 0.15, He < 0.22) in provinces such as Bursa and Muğla suggests greater genetic homogeneity, possibly due to historical bottlenecks, genetic drift, or restricted gene flow in these provinces. These factors may have led to less genetic mixing compared to Erzurum provinces. Conservation and breeding strategies should introduce genetic variation to mitigate risks associated with low diversity, such as reduced adaptability to environmental changes. Comparatively, the observed heterozygosity and expected heterozygosity in these provinces were below the range reported in other studies (Ho = 0.21-0.25, He = 0.44-0.59), which may reflect local adaptation or limited genetic exchange (Last et al., 2013). The genetic distance among accessions ranged from 0.065 (Türkiye3 and Türkiye23) to 0.257 (Kars6 and Türkiye7), with an average of 0.182, reflecting moderate genetic variability across the sampled province. This falls within the range reported for other studies using the AFLP marker on Dactylis (0.0692 to 0.4214; Peng et al., 2006), suggesting that although variability exists, some provinces may have more closely related genetic backgrounds.

The Analysis of Molecular Variance based on provinces indicated that the majority of the genetic variation (95.4%) was attributed to differences among populations while a smaller proportion (4.6%) occurred among individual accessions. This high level of inter-population differentiation is notable and exceeds findings from other studies on Dactylis germplasm using SSR markers, where within-population variation ranged from 63.3% to 74.9% (Xie et al., 2010). The greater variation among populations in this study may be influenced by the geographic isolation of some regions or ecological barriers that limit gene flow, leading to genetic drift and local adaptation.

The highest PIC value based on provincial data was recorded for Erzurum (0.21), which aligns with the findings of Sun et al. (2017), who reported similar PIC in wild D. glomerata germplasm using AFLP markers in China. However, our PIC values were lower than those reported in previous studies, such as those of Chtourou-Ghorbel et al. (2024) and Sun et al. (2017), which documented PIC values of 0.69 and 0.25, respectively. Xie et al. (2010) also reported higher PIC values (0.30 and 0.44) than ours in orchardgrass cultivars and breeding lines from North America using SSR markers. The discrepancies among these reports, including the present study, may be attributed to differences in the genetic markers used, sampling strategies, or population structure, which could influence the level of detected genetic diversity. The lower PIC values observed here (0.18-0.21) could be attributed to the bi-allelic nature of SNPs compared to multi-allelic SSR markers, which often capture a broader spectrum of genetic variation (Neuhaus and Horn, 2004). Nonetheless, SNPs offer higher resolution for genome-wide assessments and are suitable for large-scale genetic studies. Our study found greater genetic diversity in D. glomerata, with MAF, Ho, and He values surpassing those reported by Hodkinson et al. (2019), who observed a MAF of 0.05, Ho of 0.25, and He of 0.30. This increase may be due to more diverse sampling across regions, better genome coverage using GBS-based SNP markers, and potential historical admixture or outcrossing in our populations. Additionally, natural selection across varied environments could have contributed to maintaining greater genetic variability in our germplasm collection.

The observed genetic distance values and diversity patterns are consistent with previously reported findings across different Dactylis populations, where genetic differentiation often occurs due to regional environmental influences. The average genetic similarity among accessions in other studies ranged widely, from 0.43 to 0.94 (Xie et al., 2010), which supports the observed genetic variability found in this research. However, provinces with lower diversity, such as Bursa and Muğla, may require targeted conservation efforts to enhance genetic variability and reduce the risk of inbreeding. The overall fixation index (FST = 0.143) suggests moderate genetic differentiation, which aligns with findings from Madesis et al. (2014), who reported an FST of 0.186 for Dactylis, indicating that while there is genetic differentiation among populations, gene flow is not entirely restricted. Similar levels of differentiation were observed in other outcrossing grasses, where genetic diversity is primarily maintained within populations (Bolaric et al., 2005; Fjellheim and Rognli, 2005).

The clustering algorithms, UPGMA STRUCTURE, and PCoA were used to assess the genetic differentiation and grouping of the studied germplasm ( Figures 3A, B , 4 ). However, the UPGMA method provided clearer classification by further subdividing the germplasm into distinct subclusters. Therefore, UPGMA was used as the primary clustering method in this study. The UPGMA clustering divided the germplasm into three groups based on their geographical characteristics. In Cluster I, particularly in sub-cluster Ib, all accessions except for Antalya1, Samsun1, and Antalya2 exhibit traits associated with a continental harsh climate.In sub-cluster Ia, the accessions are similar in terms of climate and altitude. Cluster II: sub-cluster IIb contains mostly specimens collected from Turkey without specifying the location. In sub-cluster IIa, except for genotype Sivas3, the other accessions are similar in climate and altitude and close in distance. Cluster III: This cluster contains 40 accessions including the Iranian genotype and except for genotype Muğla2, the accessions are similar in climate and altitude (1000-1350 m).

Furthermore, we also used STRUCTURE for clustering and STRUCTURE also split the studied germplasm into three populations (Population 1, 2 and 3) geographically with few exceptions ( Figure 2A ). Such as Population 1 is the largest population and there are admixture accessions present in this population that are climatically and altitudinally similar moreover, population 2 is climatically similar and close in distance except Erzurum2. Population 3 is climatically and altitudinally similar except for Antalya1, Samsun1, Muğla3, and Antalya2. The PCoA analysis further corroborated the findings of UPGMA, dividing the germplasm into three distinct groups with a clustering pattern largely consistent with UPGMA results ( Figure 4 ). This grouping highlights the reliability of the observed genetic differentiation. The grouping reflected similarities based on geographical characteristics and dry biomass weight, a critical trait influencing species adaptability and productivity. The PCoA provided a complementary perspective by visually emphasizing the genetic relationships within and between groups, supporting the robustness of the clustering patterns observed in UPGMA. Together, these analyses highlight the strong genetic structure within the studied germplasm, offering valuable insights for targeted breeding and conservation strategies. We also evaluated the clustering pattern based on dry biomass weight, a critical characteristic of any species of forage (Lutatenekwa et al., 2020; Capstaff and Miller, 2018). The UPGMA cluster analysis based on SNP markers correlated the accessions with their production of biomass. This distinction is particularly pronounced for the high- and low-biomass materials. For instance, all productive accessions were distinctly grouped in cluster II (average dry biomass 128 g/plant), separate from the low-yield accessions in cluster I (100.7 g/plant), whereas a predominant cluster (III) comprised just samples with low biomass (83.8 g/plant) ( Figures 3A, B ). Similar associations between molecular markers-based genetic clustering and biomass productivity have been reported for other forage grass species, such as Trichloris crinita (Cavagnaro et al., 2006).

We also measured the genetic diversity indices based on UPGMA clustering. Cluster I showed moderate genetic diversity, with a mean GD of 0.22, PIC of 0.19, and MAF of 0.14. The observed and expected heterozygosities were nearly identical (0.23 and 0.2369, respectively), and the inbreeding coefficient was slightly negative (-0.03), indicates that there is a slight excess of heterozygotes, which may be a result of selection favoring heterozygous individuals, as seen in alfalfa populations where inbreeding proceeds more slowly than expected (Osborn et al., 1997). The intermediate biomass production associated with Cluster I suggests that this genetic makeup may contribute to a balance between genetic variation and trait stability. The slightly higher PIC compared to Cluster II and III implies a relatively higher potential for identifying polymorphisms in this group, possibly due to a more balanced allele distribution, which could be advantageous for breeding programs aiming to improve biomass (Sahu et al., 2020; Thavamanikumar et al., 2011).

Genetic diversity metrics in Cluster II were similar to those of Cluster I, with a GD of 0.22, PIC of 0.18, and MAF of 0.14. However, the expected heterozygosity (0.2461) was marginally higher, and the more negative inbreeding coefficient (Fi = -0.07) suggested a trend toward outbreeding. The higher heterozygosity and potential outbreeding indicate a more diverse genetic pool within Cluster II, which aligns with the higher dry biomass observed in this group. The association between genetic diversity and biomass could be attributed to heterosis (hybrid vigor), where greater genetic variation promotes better adaptation and growth (Fu et al., 2014). The outbreeding trend within this cluster may also enhance the combination of favorable alleles, contributing to increased biomass.

Cluster III is the largest cluster, including 40 accessions, and exhibited slightly lower genetic diversity metrics, with a GD of 0.21, PIC of 0.18, and observed and expected heterozygosity of 0.21 and 0.2121, respectively. The nearly zero inbreeding coefficient (-0.01) suggests balanced allele frequencies and random mating. The low biomass associated with this cluster may be related to the lower genetic diversity observed, potentially limiting the expression of advantageous traits. The reduced diversity could also indicate genetic drift or selection pressure within this group, which may have constrained the range of adaptive genetic variation and resulted in lower trait performance, such as biomass production. The AMOVA results further emphasize the genetic structure observed among the clusters, revealing that 85.8% of the genetic variation occurred within clusters, while only 14.2% was attributed to differences among clusters. This distribution suggests that the majority of genetic diversity is retained within each cluster rather than between them. The low overall FST value of 0.007 indicates minimal genetic differentiation among the clusters, implying a shared genetic background or similar evolutionary forces across the groups. The pairwise FST values between clusters also show limited differentiation, with values ranging from -0.004 between Clusters I and II to 0.015 between Clusters II and III, indicating that the clustering based on genetic similarity rather than geographic origin captures more homogeneity within the groups. Previous studies Sun et al. (2017) and Madesis et al. (2014), reported moderate differentiation in Dactylis populations according to reported FST values of 0.135 and 0.186, respectively. Overall, the results indicate that genetic diversity positively correlates with dry biomass, as evidenced by the higher values in Cluster II, which also had higher heterozygosity. This relationship highlights the importance of maintaining or enhancing genetic diversity in breeding programs to optimize key forage traits like biomass productivity.

This study indicates a novel contribution, as no previous GWAS aimed at DBW in D. glomerata has been reported. Significant MTAs were identified, with key markers on Chromosomes 1, 6, and 7, highlighting key genomic regions for biomass accumulation and valuable targets for marker-assisted breeding. Tang et al. (2018) identified 60 QTLs associated with biomass related traits in tetraploid orchardgrass, including traits like plant height, tiller number, and dry weight per plant. Their findings determined significant correlations between these traits and dry weight, providing a way for molecular-assisted breeding. Notably, their identification of QTLs for dry weight on linkage groups overlapping with Chromosome 6 aligns with our findings, highlighting shared genomic regions linked to biomass accumulation, a trait significant for forage yield improvement. Similarities in genetic loci across studies might suggest conserved genomic regions influencing biomass traits across related species, warranting further comparative genomics investigations.

The findings are particularly significant given the context of breeding programs aimed at improving forage yield. The application of GWAS based insights for marker-assisted selection could enhance the efficiency of breeding strategies, as demonstrated by the successful use of such approaches in other forage grasses like perennial buffel grass (Negawo et al., 2024). Future work should expand on these findings by integrating multi-environment trials to validate the stability of these loci under varying environmental conditions. Additionally, functional genomics approaches, such as transcriptomics and CRISPR-based validation, could elucidate the biological mechanisms linking these markers to dry biomass traits. This study paves the way for further D. glomerata genetic improvement, which could support sustainable fodder production systems.

The study provides a comprehensive assessment of the genetic diversity and population structure in D. glomerata using GBS-based SNP markers, identifying high variability across accessions and offering valuable insights for conservation and breeding efforts. The high genetic diversity observed in provinces such as Erzurum and Sivas emphasize their potential as genetic reservoirs for breeding programs aimed at improving forage yield, quality, and stress tolerance. In contrast, the lower diversity observed in provinces like Bursa and Muğla underscores the need for targeted conservation strategies to mitigate genetic homogeneity. The association between genetic diversity and biomass production, particularly in Cluster II, highlights the role of heterosis in promoting growth and adaptability. Minimal inbreeding across the accessions further suggests that genetic variability can be effectively maintained through outbreeding strategies. Moreover, GWAS identified significant marker-trait associations for dry biomass weight, a critical agronomic trait, with markers DArT-100715788, DArT-101043591, and DArT-101171265 and DArT-101090822 located on Chromosomes 1, 6, and 7 respectively. These results underline the need for conserving regional diversity while promoting gene flow across populations to optimize D. glomerata’s potential in both forage production and environmental resilience. Additionally, the genetic regions identified in this study provide valuable targets for marker-assisted selection, paving the way for advanced breeding programs aimed at improving fodder yield.