Genome-wide association study of pre-indicative agronomic traits affecting malt quality in Ethiopian barley (Hordeum vulgare L.) germplasm

Abstract

Background

Barley (Hordeum vulgare L.) is a diploid cereal crop and one of the major crops grown in Ethiopia, with a long history of domestication. It serves two primary purposes: as a staple food and as a raw material to produce alcohol. Barley germplasm resources can serve as sources of new alleles in breeding programs. This study aimed to evaluate the performance of diverse germplasm for primary malt quality indicators, agro-morphological and nutritional traits across diverse environments and to identify marker-trait associations with these traits using genome-wide studies (GWAS) among 260 barley germplasm. The experiment was conducted at four experimental sites, Holeta, Debre Markos, Bekoji, and Welkite Agricultural Research Centers using an Alpha Lattice Design. Nine agronomic and nutritional data were collected on Heading Date (HD), Flowering Date (FD), Maturity Date (MD), Plant Height (PH), Pedicle Length (PL), Grain Size (GS), Free Amino Nitrogen (FAN), Soluble Nitrogen (SN) and Moisture Content (MC).

Results

Among the 260 genotypes, 43% had straw-colored glumes, 33.4% had black, 15.7% had purple, and 7.6% had light-colored glumes. Additionally, 72.2% had two rows, 19.6% had six rows, and 6.1% had irregular rows. The narrow-sense heritability estimates are based on quantitative data, ranging from 0.52 to 0.97. Correlation analysis revealed that FAN have a strong positive correlation with SN (r = 0.91), HD has a negative correlation with MD (r² = -0.03), MC has a weak positive correlation with all traits (r² = 0.06–0.14), and FD has a moderate correlation with PL (r² = 0.43). Principal Component Analysis (PCA) identified three major components: PCA1 (39.13%), PCA2 (22.01%), and PCA3 (19.28%). The population were clustered into three k-means Cluster 1 (31 accessions s), Cluster 2 (107 accessions), and Cluster 3 (122 accessions). A total of > 50 Mb SNP density was generated across seven barley chromosomes. Linkage disequilibrium decay is 177.76 kb, with an r² value of 0.2. In total, 65 SNP markers were significantly associated with HD, FD, MD, PH, PL, GC, FAN, and SN. These include 2 SNPs for HD (5 H), 7 SNPs for FD (2 H, 5 H, 6 H, 7 H), 13 SNPs for MD (2 H, 3 H, 7 H), 3 SNPs for PH (2 H, 5 H), 3 SNPs for PL (5 H, 6 H),23 SNPs for GC(1 H, 2 H, 4 H, 5 H, 6 H and 7 H), 5 SNPs for FAN(1 H) and 7 SNPs for SN(1 H, 2 H, 4 H, 5 H and 7 H), with a P-value range of 3.0–5.8. The minor allele frequencies range from 0.24 to 0.52.

Conclusion

The identification of malt quality pre-indicative agronomic traits under natural conditions lays a foundation for malt barley improvement. To ensure industrial relevance, these traits and associated SNP markers must be validated under real-world processing conditions for their application in molecular breeding. This will drive precision breeding strategies, enhancing malt quality, market competitiveness, and industrial sustainability in Ethiopia.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12555-8.

Introduction

Barley (Hordeum vulgare L.) is a diploid and one of the major cereal crops grown in Ethiopia [1, 2]. It is the fourth most important staple food crop in the world, after wheat, rice, and maize [3, 4]. It is also one of the oldest domesticated crops, having been first cultivated in the Fertile Crescent around 10,000 years ago [5–8]. Barley is a staple food in many regions due to its adaptability to high altitudes, drought, and soil salinity [9, 10]. Both husked and naked barley varieties are used for animal feed, malt, and food. Additionally, barley is widely grown for its health benefits in most parts of the world [2, 11].

The world’s top barley-producing countries are the European Union, Russia, and Australia, with total production volumes of 51.5, 21.5, and 13.5 million metric tons, respectively [12]. Ethiopia is the second-largest barley-producing country in Africa, contributing 34.5% of the continent’s total barley production [13]. In Ethiopia, barley is grown at altitudes between 1,800 and 3,400 m above sea level [14]. The barley-producing belt in Ethiopia covers 799, 127.84 hectares, yielding approximately 2.04 million metric tons of barley annually. However, despite this, Ethiopia’s barley production remains low compared to global production levels [15].

Barley is grown in Ethiopia and elsewhere for food, feed, and malt [16], with malt barley being the most commonly used raw material for brewing beer and for distillation. Although both 2-row and 6-row barley are used for malting, 2-row barley is preferred, while 6-row barley is primarily used for food, both globally and in Ethiopia [17, 18]. The Barley Breeding Program has two primary objectives: to develop high-protein, superior-quality barley for food and feed applications, and to produce barley varieties with low protein and high sugar content for specific end-use requirements [16]. In addition to efforts to increase yield, enhancing malt quality remains a major research focus in malting barley breeding, despite the complexity and challenges it presents [16, 19, 20]. Malt quality traits are complex because they are influenced by processes from the field to the malting factory, and most malt quality indicators are polygenic, controlled by multiple genes [16]. Malt quality traits include high malt extract, low protein content, good solubility characteristics, good kernel formation, low glume content, and high, uniform grain size [16, 21].

Malt quality can be evaluated based on natural field conditions and processing performance in a malt factory [22–24]. In natural field conditions, several agronomic traits influence malt barley quality [25]. Glume Colour: Lighter glume Colours (golden, straw) are preferred due to lower polyphenol content, reducing astringency and improving beer stability [26–29]. Darker Colours (black, purple) increase polyphenols, enzyme activity, and beer haze [29, 30]. Heading Date: Early heading is desirable, as late heading increases susceptibility to drought, heat stress, and fungal infections, reducing starch accumulation and malt quality [31–34]. Flowering Date: Earlier flowering enhances sugar accumulation, while late flowering leads to higher protein content and lower malt extract efficiency [31, 35–37]. Maturity Date: Early maturity yields plumper grains with higher starch content, which is essential for malt extract production. Delayed maturity negatively impacts enzymatic conversion and storage viability [36, 37]. Plant Height: Shorter or semi-dwarf varieties provide more uniform grain size and higher malt extract potential [38]. Taller plants are prone to lodging, fungal contamination, and inconsistent germination. Peduncle Length: A longer peduncle improves air circulation, reducing disease risk (e.g., Fusarium head blight) and enhancing nutrient transfer [39]. However, excessive length may slightly delay grain maturity [40]. Optimizing these traits enhances malt quality, ensuring better extract yield, stability, and brewing performance [41]. Grain size, free amino nitrogen (FAN), soluble nitrogen (SN), and moisture content (MC) are key factors in malt quality, affecting both brewing efficiency and beer characteristics. Bigger, consistent kernels enhance steeping, germination, and yield of extracts [42, 43]. FAN, originating from protein breakdown, supplies crucial nitrogen for yeast; however, both a lack and an excess can hinder fermentation and stability [44, 45]. SN signifies protein alteration, influencing clarity, foam, and flavor, but imbalances diminish effectiveness [46, 47]. MC is essential for enzyme activity and storage, where low concentrations improve shelf life while high concentrations elevate microbial hazards [24, 48]. Together, these characteristics establish malt suitability and need to be optimized to satisfy brewing standards [49].

In Ethiopia, malting barley varieties have been largely developed through conventional breeding, which, while effective, is time-consuming and heavily dependent on environmental conditions [50, 51]. This limits the efficiency of selecting, such as grain quality, and delays the release of improved varieties [52]. As a center of barley diversity, Ethiopia holds valuable landraces with great potential for crop improvement [53, 54]. Integrating modern molecular tools, such as Genome-wide association study (GWAS), can accelerate breeding, enhance precision in trait selection, and unlock this genetic potential [55]. This approach can strengthen the malt barley value chain and meet the growing needs of Ethiopia’s brewing and malting industries [54, 56].

There are various ancient and modern methods for identifying malting barley traits, including selection and microsatellite SSR markers. GWAS is one of the modern molecular genomic tools widely used in crop breeding programs [57–59]. In the present study, GWAS is performed by associating single-nucleotide polymorphisms (SNPs) with natural populations for malting quality-indicating traits. With current high-throughput next-generation sequencing technology, SNP data can be easily generated, significantly enhancing their utility for crop improvement [60–62]. SNP accession data are now readily available, enabling GWAS to run much faster and more efficiently. Diverse populations, high-density markers, multi-environmental trials, and polymorphic SNP markers are used for GWAS-based studies [61, 63, 64]. Identifying the genetic basis of agronomic traits in crop improvement is a challenging task [59]. Most important agronomic traits are polygenic and controlled by two or more genes [65]. One molecular tool that has become increasingly popular and powerful in plant genetics research over the past decade is the GWAS [66, 67]. Random mating and subsequent selection have led to highly structured populations for barley breeding and improvement [68]. Several studies in Ethiopia and globally have aimed to uncover the genetic basis of malting quality traits in barley. Notably, Sintayehu Daba et al. [55]. applied GWAS using a panel of Ethiopian and USA landraces and breeding lines, identifying SNPs linked to key malting traits and highlighting the genetic richness of local germplasm. Similarly, Yunxia Fang et al. [47]. and Mark E. Looseley et al. [69]. conducted GWAS in diverse populations, emphasizing the value of high-throughput genotyping and the inclusion of landraces and wild relatives. Collectively, these studies show that integrating genomic tools with diverse germplasm is a powerful approach to accelerate the development of superior malting barley varieties. Therefore, this study aimed to evaluate the primary malt quality indicators of Ethiopian barley germplasm in the field by assessing the performance of key agronomic traits and nutritional content across diverse environments, and by identifying polymorphic SNP markers associated with these traits using GWAS. This approach enhances understanding of the genetic basis of essential malting characteristics, aiding in selecting superior malt barley variability.

Materials and methods

Experimental association panel

A total of 260 Ethiopian barley accessions were used in this study. These included 21 breeding lines obtained from the Barley Improvement Program of Holeta Agricultural Research Centre and 239 landraces sourced from the Ethiopian Biodiversity Institute. The collection was established to represent the major agro-ecological zones and geographical regions of Ethiopia. In addition, the mapping panels were designed to contain 51 six-row, 193 two-row, and 16 irregular-row barley Core Collections, as well as the material utilized by Teklemariam et al. [70], who focused on drought-tolerance screening under both field and growth-chamber conditions. In contrast, the current study was designed to investigate pre-indicative traits related to malt quality across different experimental sites, independent of those used in Teklemariam’s work. The passport data for all the materials used in this study were provided as supplementary data (Supplementary file. 1). The geographical locations show that the site lies within the political boundaries of the Oromia, Amhara, and Southern Nations, Nationalities, and Peoples’ regions.

Description of study sites

Four experimental sites located within Ethiopia’s major barley-growing belt were selected for this study. The experimental site represents diverse agro-ecological zones. The four sites are Holeta Agricultural Research Centre, Kulumsa Agricultural Research Centre (Bekoji Sub-Centre), Welkite Agricultural Research Centre (Jenbero Sub-Centre), and Debre Markos Agricultural Research Centre. A full description of the experimental site is provided in Table 1. These sites were chosen to capture the environmental variability within the barley belt of Ethiopia, ensuring robust evaluation of genotypic performance under diverse growing conditions for malt quality pre-indicative traits (Fig. 1).

Table 1.

Geographic, climatic, and soil characteristics of the experimental sites used in this study

ES	Latitude	Longitude	Altitude (M)	Aarf (mm)	MaT	MiT	pH	ST
HARC	9°00’ N	38° 30’ E	2,400	1,044	20.4 °C	13.8 °C	5.23	Nitosols
WARC-JSC	7° 56’ N	37°51’ E	2,219	1, 274.67	27.5 °C	16.5 °C	5.63	clay
DMARC	10° 16’ N	37° 46’ E	2,467	1,340	19.5 °C	9.9 °C	5.72	clay
KARC-BSC	7°35′N	39°10′E	2,810	939	27.5 °C	16.5 °C	5.63	clay

HARC Holeta Agricultural Research Centre, WARC-JS Welkite Agricultural Research Centre -Jenbero Sub-Centre, DMARC Debre Markos Agricultural Research Centre, KARC Kulumsa Agricultural Research Centre - Bekoji Sub-Centre, ES Experimental site, Aarf Average Annual Rainfall, MaT Maximum Temperature, MiT Minimum Temperature, ST Soil Type, M Meter, mm millimeter

Fig. 1.

A map illustrates the locations of the four experimental sites where the research trials were conducted. The map provides a visual representation of the study areas, highlighting their geographical distribution and significance to the experiment. Note: Oromia (blue), Amhara (red), and SNNPR (yellow) are Ethiopian regional states (political demarcation) where the experimental sites are located

Experimental design

An alpha lattice design with two replications was used for this experiment in the 2024/2025 cropping season. The experiment was conducted within a single year to ensure uniform management and data collection while minimizing inter-annual environmental variation. This approach enabled a clear assessment of genotype performance across diverse environments for reliable GWAS analysis. The full design includes a total of 13 blocks and 20 plots in each block, the plot size being 1 m in length and 0.8-meter in width. Each plot has four rows, and each row in turn is drawn by a row marker material. A spacing of 1 m was maintained between blocks, 50 cm between plots, and 20 cm between rows within each plot. Each association panel was assigned to each experimental plot in a randomized manner. The randomization was done using a random allocation software for parallel-group randomized trials [71].

The experimental accessions were planted on different days using the same design for all experimental sites. Before sowing, all experimental sites were prepared with the necessary land preparation. The experiment was sown by weighing 10 g of seed from each association panel barley accession for one plot, and the total seed for each row was 2.5 g; equal amounts of seed were sown for the four rows in the plot. 45 kg of DAP (Di-ammonium phosphate) soil fertilizer per hectare was applied during sowing. Weeds were manually weeded once a month for 4 consecutive months in all four experimental sites. All four experimental sites were managed without pesticide application. Sentries were assigned to prevent damage by birds and animals, and data was collected intact from all sites.

Phenotype data collection

The International Plant Genetic Resources barley descriptor (Descriptors for Barley: Hordeum Vulgare L.) [72] Served as the basis for the agronomic data collected for this research investigation. Qualitative and quantitative data were collected using both plot-based and plant-based agronomic methods. Five plants were randomly selected from the two middle rows of a plot’s four rows and marked with permanent markers for the plant-based data. The quantitative data collected in this study included plant height (PH) and peduncle length (PL). Plant height was measured at maturity, in centimeters, from the ground to the tip of the spike. Peduncle length was measured in centimeters from the last leaf node to the base of the spike.

Plot-based data included heading date (HD), flowering date (FD), maturity date (MD), glume colour (GLC), and row number (RN). While HD was recorded as the number of days from sowing to the date when 50% of the heads in a plot had emerged, FD was measured as the number of days from sowing to when 75% of the flowers had fully flowered. On the other hand, MD was recorded as the number of days from sowing to when 75% of the plants in a plot had matured and exhibited a colour change indicative of physiological maturity. After harvesting, all genotypes were subjected to nutritional data, which was also collected as genotype-based data GS was measured in millimeter, FAN was measured in milligram, SN was measured in milligram and MC was measured in grams.

The qualitative data included GLC and RN, although the recording was based on the plot. Glume colour was scored based on visual observation of the glumes emerging in each plot and classified into four categories: straw, purple, black, and yellow. The row number referred to the arrangement of rows per spike and was categorized as two-rowed, six-rowed, or irregular.

Barely grain size and nutrition data

For barley grain size, a 100-gramme grain sample was sieved using a set of sieves with openings of > 2.8 mm, > 2.5 mm, > 2.2 mm and < 2.2 mm. Sieving was carried out for five minutes, after which the grains were separated into four different size categories. For this study, the weight of only the sieves at > 2.2 mm was used and subjected to GWAS analysis. Whole barley grain samples (100 g per sample) were cleaned and equilibrated at room temperature to standardize moisture content across all genotypes. Samples were scanned directly using a near-infrared spectroscopy (NIRS) instrument equipped with a rotating beaker to ensure uniform visualization without compaction. Spectral data was collected in the form of 32 cumulative scans per replicate in the 1250 nm range, with reference scans performed before each batch. The final data set included free amino nitrogen and soluble nitrogen. The moisture content for each sample was determined by weighing 100 g of grain kernels (initial weight) and drying them in an oven at 105 °C for 16 h. After drying, the samples were weighed, and the moisture content was calculated using the following formula:

Genotyping

Following Teklemariam et al. [73]., for genotyping, three seeds per accession of the association panel (260 in total) were grown under controlled greenhouse conditions (20–22 °C day/17–19 °C night) in nutrient-rich Einheitserde ED73 soil. At the 2–3 leaf stage, approximately 300 mg of leaf tissue from a single plant was sampled for DNA extraction using a modified CTAB (Cetyltrimethyle Ammonium Bromide) protocol. Genotyping was conducted using the barley 50k iSelect SNP array (Illumina) at Trait Genetics GmbH, Germany.

From an initial set of 40,387 SNPs, 10,644 high-quality markers were retained after removing monomorphic loci and applying filtering thresholds for missing data (≤ 5%), minor allele frequency (≥ 5%), and heterozygosity (≤ 12.5%). These 10,644 high-density SNPs were used for all subsequent analyses in this study (Supplementary File 2).

Statistical analysis

Phenotypic data analysis

The malt quality, as indicated by pre-indicative agronomic and nutritional data generated for this study, underwent statistical analysis to assess the genetic variability of the mapping panel and evaluate the environmental impact on its performance.

Pie chart visualization

A pie chart was used to visualize qualitative data, including GLC and RN. The data visualization was performed using R software (version 4.5) [74], with the ggplot2 (package (ggplot2., 2025) [75], utilized to create and visualized a 3D pie chart.

Correlation analysis

The collected agronomic parameter data were tested for normality prior to analysis using the R software (version 4.5) [74], qqPlot packages [76, 77]. Normal was assessed using distribution curves (Supplementary file. 3) and Quintile- Quintile (Q-Q) plots (Supplementary File. 4), with statistical significance evaluated at P ≤ 0.05. The normally distributed agronomic data was entered into the correlation analysis. The person’s correlation coefficients analyses were constructed for the agronomic traits using the R software (version 4.5) [74], Core package. The correlation analysis was performed using a P ≤ 0.05, mainly designed to show the level of correlation between scored agronomic data.

K-means and hierarchical clustering

The generated phenotype and nutritional data were analyzed using the statistical software JMP [78], used for K-Means clustering was performed to categorize the genotypes into different groups based on their similarity. Hierarchical clustering analysis was conducted based on malt quality-related agronomic traits. Euclidean distance was calculated to construct a dissimilarity matrix, which was then used to perform hierarchical clustering using the Unweighted Pair Group Method with Arithmetic Mean (UPGMA). The results were visualized using a dendrogram with different colours based on their similarity, providing a clear representation of trait-based groupings. To visualize multivariate patterns, a parallel coordinate plot was created. This analysis was carried out using the cluster package [79] in R software (version 4.5) [74].

Principal component analysis (PCA)

The circular PCA analysis was conducted to assess the population structure of the mapping panel, along with the four experimental sites. This analysis aimed to determine the genetic variability within the mapping panel. The analysis was performed using R software (version 4.5) [74], utilizing the PCA tools package [80]. A significance threshold of p ≤ 0.05 was applied, with a Bonferroni correction to account for multiple testing. A 3D scatterplot was then constructed to illustrate group separation across the main dimensions.

Genetic admixture analysis

To further understand the genetic composition and population structure of the mapping panel, genetic admixture analysis was performed. The aim was to determine the degree of shared ancestry among barley accessions and to assess how genetic components were distributed across their geographic collection regions. Admixture analysis can reveal historical gene flow, breeding patterns, and potential subpopulations within the panel. Genetic admixture analysis was conducted using the LEA (Landscape and Ecological Association Studies) package [81], which implements sparse non-negative matrix factorization (sNMF) a fast and efficient algorithm for estimating individual ancestry coefficients and detecting population structure. Multiple values of the number of ancestral populations (K) were tested, and cross-entropy scores were used to determine the optimal K. Bar plots were generated to visualize individual admixture proportions, highlighting the genetic background and subpopulation structure of each accession.

Genotypic data analysis

The genotype data generated from the mapping panel accessions was subjected to two types of analysis: Linkage Disequilibrium Decay (LDD) and GWAS. These analyses were intended to assess the genetic variability of SNPs within the mapping panel and to identify significant SNP markers strongly associated with malt quality pre-indicative agronomic traits SNP genetic variability within the mapping panel and to identify significant SNP markers strongly associated with pre-indicative agronomic traits indicative of malt quality.

Linkage disequilibrium decay (LDD) analysis

The LDD analysis was conducted by calculating pairwise linkage disequilibrium (r²) between SNPs across the genome using SNP data. SNP pairs were grouped based on physical distance, and average r² values were computed for each distance bin. A smooth curve was then fitted to the SNP data to model how LDD is with increasing distance. The decay distance was defined as the point at which the fitted curve intersects a threshold of r² = 0.2, indicating the extent of linkage across the genome. This analysis provides insight into the recombination landscape and SNP marker resolution in the studied population. We used the coefficient of determination (r²) to assess SNP associations. The formula for the Coefficient of determination (r²) is as follows:

Where N represents the effective population size, and C is the recombination rate between loci. Based on the formula, the results were interpreted as follows: r² = 1 indicates complete linkage disequilibrium (LD), meaning the alleles are perfectly correlated. r² > 0.8 suggests high LD, which is useful for selecting tag SNPs, while r² < 0.2 indicates low LD, meaning the loci are mostly independent. The analysis was conducted using R software (version 4.5) [74] using LDheatmap R package [82].

Genome wide association study (GWAS)

The R program was used for phenotypic analysis [48]. Tukey’s technique was testing the outliers, both within and between experiments, outliers were eliminated. Equation 1 indicated below were used to estimate trait’s narrow sense heritability (h²) (Supplementary File. 5), using the Linear Mixed-Effects Models for Heritability (lme4) R package [83]. Equations 2 and 3 below were used to estimate repeatability values (R). Using the Restricted Maximum Likelihood (REML) approach, Meta R software was used to compute the best linear unbiased Predictor (BLUP) for each trait based on the linear model described by the Eq. 4 [84]. BLUP is recommended for large and diverse populations, as it treats genotypes as random effects, minimizes random variation or errors in the data through shrinkage towards the population mean, and provides more accurate genetic performance across environments. The variance components of the genotypes, genotypes with experiment interactions and the residuals were represented by (additive genetic variance), (total phenotypic variance, including additive, dominance, genotype × environment interaction, and residual/error variance), (Additive genetic variance; The portion of total phenotypic variance due to the sum of additive effects of alleles). The number of biological replicates is nR, and the number of experiments is n_o.

1

2

3

4

The R package Genome Association and Prediction Integrated Tool (GAPIT) was used for the genome-wide association analysis [85]. The FarmCPU model was used for GWAS analysis [86]. The FarmCPU model is explained as follows:

Where Y denotes the Phenotype vector, X is the Genotypic data matrix for the markers, β stands for Effect sizes of SNP markers, M_j​ is the genotype of the j^th marker being tested, α_j​ is the effect of the j^th marker, and ε∼N (0,σ²I) is the residual error term. The GWAS was conducted for each experimental site, and common SNPs were identified as stable markers. To select significant markers, a threshold of a logarithm of the odds (LOD) score of 3 or higher was applied. This approach ensured that only those SNPs with strong statistical support were considered, while minimizing the potential for false positives. The common SNPs across all experimental sites were prioritized as they represent stable markers that are less influenced by site-specific environmental or genetic variations, providing greater reliability for further genetic analysis and marker development. The Manhattan plot was constructed using R software (version 4.5) [74] using the CMplot package [87].

Results

Evaluation phenotype data

In this study, genetic variability associated with pre-indicative traits of malt quality was evaluated using phenotypic and nutritional data collected across four experimental sites to assess both genetic and environmental influences. Two complementary approaches were used. First, the mapping panel was grouped based on two qualitative traits - glume colour and row type - to investigate their possible association with malt quality. These traits, which are often associated with morphological and historical selection differences, can directly influence grain characteristics. Secondly, 9 quantitative traits known to affect malt quality, namely maturity date, flowering date, ripening date, plant height, stalk length, grain size, free amino nitrogen, soluble nitrogen and moisture content, were assessed by phenotypic analysis. These traits are related to plant development, grain filling and nutrient content, which are critical for grain quality and malting performance. This combined approach enabled the identification of genetic variation and genotype-by-environment interactions that affect malt production efficiency. The results provide useful insights and genetic resources for breeding programs aimed at improving malt quality through marker-assisted selection.

Qualitative phenotype data results

Among the accessions evaluated, glume colour was categorized into straw, black, purple, and yellow. Most accessions exhibited straw-colored glumes, followed by black and purple. The distribution of Glume Colours presented in Fig. 2, indicating SC as the most prevalent category, accounting for 43% of the mapping panels. BC (33.4%), PC (15.7%), and YC (7.6%) accounted for the glume colour distribution, respectively. This distribution suggests a dominance of certain glume types across the barley population. These findings suggest that glume colour, while a qualitative trait, may serve as a useful pre-indicative marker for selecting accessions with favorable malt quality attributes, particularly in early breeding cycles where complex trait phenotyping is impractical. These results revealed that most of the glumes exhibited straw and yellow colours, which are considered pre-indicative traits of malt quality. In contrast, black and purple glume colours are identified as unacceptable malt quality pre-indicative traits. Based on these findings, 50.6% of the mapping panels are believed acceptable for malt quality, while 49.4% are not.

Fig. 2.

Percentage composition and distribution of glume colour in the barley mapping panel: 43% SC (Straw Colour), 33.4% BC (Black Colour), 15.7% PC (Purple Colour), and 7.6% YC (Yellow Colour)

Similarly, the Barley Row Number analysis, as illustrated in Fig. 3, shows that most of the mapping panels belong to the TR category, accounting for 74.2%. This is followed by the SR category at 19.6% and the IR category at 6.1%. The results indicate that the two-row barley type is preferred for malt quality, as it is most associated with superior malting characteristics. In contrast, the lower proportions of six-row and irregular-row barley suggest that these types are less suitable for high-quality malt production.

Fig. 3.

Percentage composition and distribution of the two-row and six-row barley germplasm in the mapping population’s; 74.2% TR (two row), 19.6% SR (six row), and 6.1% IR (irregular row)

Quantitative phenotype data

The nine quantitative traits - HD, FD, MD, PH, PL, GS, FAN, SN and MC, were subjected to quantitative trait statistical analyses, yielding distinct results. The statistical analyses performed are summarized below:

K-means hierarchical clustering

The K-means clustering grouped the association panel into three distinct clusters: Cluster 1 contained 31 genotypes, Cluster 2 contained 107 genotypes, and Cluster 3 contained 122 genotypes (Table 2). The assignment of each genotype to its respective cluster is provided in Supplementary File 6. The barley genotypes’ differences in HD, FD, MD, PH, PL, GS, FAN, SN and MC are displayed in the parallel coordinate plot. The association panel grouped into three clusters based on agronomic and malt quality pre-indicative traits at a significant level of p ≤ 0.05 (Fig. 4). Three clusters of genotypes (red, green, and blue) are formed, and mean trends are indicated by bold gray lines. HD and MD stay steady, MC exhibits mild convergence, and PL and FAN show the most disparities, which drive cluster separation. The greatest discriminating characteristic is FAN. The dendrogram uses hierarchical clustering to show the links between genes and the association panel, including both agronomic and nutritional data (Fig. 5). Different genotype groups are indicated by colored branches, and dissimilarity is reflected in branch length. Higher linkage distances cause major clusters to merge, revealing more extensive links, whereas lower distances highlight subgroups. This is an example of genotype variety, which helps breed and improve the quality of malt in Ethiopia.

Table 2.

Presents the cluster count of the mapping panel and the percentage contribution of agronomic traits within each cluster

Cluster	Counts	HD	FD	MD	PH	PL	GS	FAN	SN	MC
1	31	53.85	72.04	117.02	105.78	15.12	113.3	49.63	70.04	115.7
2	107	61.79	87.26	111.13	99.48	14.13	91.84	51.79	71.32	89.3
3	122	69.53	85.23	122.46	78.43	14.11	83.43	93.53	59.34	63.16

Fig. 4.

Parallel coordinate plot of barley genotypes grouped into three clusters based on agronomic and malt quality pre-indicative traits at a significance level of P = 0.05. The colors indicate different clusters: genotypes in Cluster 1 are shown by red lines, Cluster 2 by green lines, and Cluster 3 by blue lines. The mean trend across genotypes within each cluster is represented by the bold grey line. The lower-right panel shows an overlay comparison of the three clusters

Fig. 5.

Hierarchical clustering dendrogram of the barley association panel based on agronomic traits, nutritional content, malt quality, and pre-indicative traits at a significant level of P = 0.05. Note: Different colors indicate distinct clusters within the association panel based on trait variation. Colored dendrogram branches represent distinct genotype clusters (red, green, blue, yellow, purple, etc.). The x-axis shows individual barley genotypes, while the y-axis (linkage distance) indicates the degree of dissimilarity among clusters. Short branch lengths reflect high genetic similarity, whereas longer branches indicate greater dissimilarity

Correlation analysis

The circular correlation results (Fig. 6) indicate that the five traits exhibit varying correlation coefficients (r² values). HD showed a negative correlation with MD (r² = −0.06), suggesting that as HD increases, MD decreases, and vice versa. No correlation was observed between MD and FD (r² = 0). The strongest positive correlation was found between PH and PL (r² = 0.6), indicating that increases in PL are associated with increases in PH. Other traits exhibited weak positive correlations, with positive r² values.

Fig. 6.

Correlation between the malt quality traits and pre-indicative agro-phenological traits of the mapping panels. Legend: HD = heading day; FD = flowering day; MD = maturity day; PH = Plant height, PL = peduncle length; GS = Grain Size; FAN = Free Amino Nitrogen; SN = Soluble Nitrogen and MC = Moisture Content

Principal component analysis (PCA)

The results from the PCA analysis and the explained correlations are shown in Fig. 7. The variance explained by each component is as follows: The first principal component (PC1) accounts for approximately 39.13% of the total variance. The second and third components explain 22.01% and 19.28%, respectively. Together, the first three components account for over 80% of the total variance. The standard deviation of the components is as follows: PC1, PC2, PC3, PC4, and PC5 are 1.3987, 1.0490, 0.9818, 0.7541, and 0.64079, respectively. The proportion of variance for each component is 0.3913, 0.2201, 0.1928, 0.1138, and 0.08212, respectively. The cumulative proportion of variance is 0.3913, 0.6113, 0.8041, 0.9179, and 1.00000, respectively. The PCA loadings for each trait (HD, FD, MD, PH, PL, GS, FAN, SN and MC) vary across the components. HD, PH, GS, FAN and SN have strong loadings on the first component, while PL and MC have significant loading on the third component. Strong loadings of HD, PH, GS, FAN and SN on PC1 suggest these five traits contribute most significantly to the variation explained by the first principal component. In other words, the first component is heavily influenced by variations in HD, PH, GS, FAN and SN. This suggests that these traits are closely related, and together they account for a large portion of the total variance observed in the data. On the other hand, Significant loading of PL and MC on PC3 indicates that PL and MC are more closely associated with the third principal component. While HD, PH, GS, FAN and SN dominate the first component, PL and MC are relatively more important in explaining the variation captured by the third component. This suggests that PL and MC behave differently from HD, PH, GS, FAN and SN in terms of how they contribute to the overall variation in the dataset. The variation described in the seven-component PCA scree plot decreases sharply from PC1 to PC3, then levels out gradually from PC4 to PC7 (Fig. 8). The obvious “elbow” surrounding PC3 suggests that the first two to three components account for most of the significant fluctuation, whereas the remaining components may be noise or give little extra information, which supports their rejection from dimensionality reduction.

Fig. 7.

Principal component analysis illustrating the relationships among agronomic traits (HD, FD, MD, PH, and PL) and the distribution of accessions across the PCs at a significance level of P = 0.05

Fig. 8.

Scree plot illustrating the proportion of variance explained by the first seven principal components (PC1-PC7) in the PCA, highlighting the ‘elbow’ for dimensionality reduction. The x-axis represents the principal component number (PC1-PC7), and the y-axis represents eigenvalue (or proportion of variance explained)

Genetic admixture analysis

This study reveals the genetic differentiation and population structure of ten Ethiopian regions using pairwise F < sub > ST</sub > and structure analyses. The F < sub > ST</sub > heatmap indicates substantial divergence in regions like Gamo-Gofa (~ 3.77) and Arsi-Bale (~ 3.19), suggesting limited gene flow or historical isolation. In contrast, regions such as Hararghe, HARC, and Gojam show lower F < sub > ST</sub > values, reflecting greater genetic similarity and connectivity. Complementary structure analysis (K = 6) reveals complex admixture patterns, with some populations exhibiting dominant ancestry from a single cluster and others displaying significant genetic mixing. These results highlight varying evolutionary trajectories, ecological adaptations, and possibly localized breeding practices across regions. Together, the findings underscore the importance of incorporating population structure and genetic differentiation in conservation planning, association mapping, and the development of targeted breeding strategies aimed at preserving and utilizing the rich genetic diversity within Ethiopian germplasm (Fig. 9).

Fig. 9.

Admixture analysis revealing the genetic diversity and ancestral lineage composition of Ethiopian barley genotypes across different geographic regions and subpopulations. (A) pairwise genetic differentiation displaying the admixture patterns of the mapping panel grouped by collection region, highlighting the extent of shared ancestry and regional differentiation. (B) Structure bar plot showing the genetic structure of four inferred subpopulations, illustrating distinct clusters and varying degrees of admixture among accessions. (C) Individual-level structure bar plot showing admixture profiles for each accession and the proportion of inferred ancestral contributions. Collectively, these results demonstrate a complex genetic background within the Ethiopian barley gene pool, shaped by both regional adaptation and historical gene flow

Linkage disequilibrium decay (LDD)

The genome-wide LDD analysis demonstrated that LD, measured by the squared correlation coefficient (r²), declines with increasing physical distance between SNPs. The average r² dropped below the commonly used threshold of 0.2 at approximately 177.76 kilobases (kb), indicating that SNPs separated by distances greater than this are generally not in strong linkage. This suggests a moderate historical recombination rate and a considerable level of genetic diversity within the barley population studied. The relatively slow decay of LDD implies the presence of extended haplotype blocks, potentially shaped by factors such as population structure, historical breeding practices, and selective pressures. This LD decay distance is crucial for determining GWAS resolution, as it defines the approximate genomic window around significant SNPs and informs SNP detection confidence intervals. Moreover, it provides essential guidance on implementing marker-assisted selection (MAS), optimizing marker density in genomic selection (GS), and interpreting linkage patterns in barley improvement programs. High-density SNP genotyping across the seven barley chromosomes was achieved using the 50 K iSelect SNP array, yielding an initial set of 43,461 scorable SNP markers. After stringent quality control based on missing data rates, minor allele frequency (MAF), heterozygosity, and imputation accuracy, 10,644 high-quality SNPs were retained for downstream analysis. These markers were unevenly distributed across the barley genome, with the highest number on chromosome 2 H (1,857 SNPs), followed by 5 H (1,837), 7 H (1,695), 6 H (1,388), 3 H (1,376), 1 H (1,317), and the lowest on chromosome 4 H (1,174) (Fig. 8). This distribution reflects chromosome-specific variation in marker density and potential recombination landscapes (Supplementary File 7). Hierarchical clustering of marker trait associations based on LD further revealed genomic regions characterized by strong LD blocks (Supplementary File 8).

Genome wide association study (GWAS)

The population structure patterns underlying the marker-trait associations identified through GWAS reveal significant genetic differentiation among subgroups, which may influence association signals (Fig. 10). The GWAS results, illustrated in (Fig. 11), were obtained using the FarmCPU association model. Significant SNP-marker trait associations are identified, as detailed in Table 3. The GWAS results show LOD values ranging from a minimum of 3.0 to a maximum of 5.8, while the threshold value varies, with a minimum and maximum threshold of 3.5, respectively. From the GWAS results, the Bekoji experimental site recorded the highest number of significant marker-trait associations (MTAs), with 118 significant SNPs. This was followed by Holeta with 98 significant SNPs, Debre Markos with 93 and Jenbero with 72 significant SNP markers. Additionally, 65 SNP markers were commonly identified across all four experimental sites. The Manhattan plot displays the distribution of SNPs across the genome, highlighting significant marker-trait associations based on their p-values. Peaks in the Manhattan plot represent genomic regions with strong associations, which may contain candidate genes influencing malt quality traits. Among the nine quantitative traits analyzed, no significant SNPs were detected for MC. This suggests that the genetic basis of MC may be governed by small-effect loci or strong environmental influences, making it difficult to identify major associations under the current analysis conditions. The 3D Principal Component Analysis (PCA) plot derived from GWAS data, capturing the major axes of genetic variation among genotypes. The Q-Q plot associated with the Manhattan plot demonstrates a good fit, indicating reliable marker trait associations.

Fig. 10.

Illustration of linkage disequilibrium and population structure patterns underlying marker–trait associations identified through GWAS. (A) Three-dimensional principal component analysis (PCA) plot derived from GWAS data, representing the major axes of genetic variation among genotypes. (B) K-means clustering based on principal components, showing the grouping of accessions according to genetic similarity. Together, these analyses provide insights into population structure and LD patterns that influence GWAS results

Fig. 11.

GWAS manhattan and Q-Q plots for malt quality pre-indicative traits. Significant SNPs are indicated above the bonferroni-corrected threshold (LOD ≥ 3). Legend: Q-Q plot (Quantile-Quantile plot)

Table 3.

Presents the GWAS findings and SNP identification for malt quality pre-indicative traits, highlighting significant marker-trait associations and their potential role in influencing malt quality characteristics

No	Traits	Chromosome	LOD	MAF	Position(bp)	SNP Markers	NMTA	SP (bp)	EP (bp)
		2 H	3.653	0.341	689,512,037	JHI-Hv50k-2016–114427		689,334,277	689,689,797
1	PH	2 H	3.949	0.46	689,273,079	JHI-Hv50k-2016–114347	3	689,095,319	689,450,839
		5 H	3.282	0.52	569,020,612	JHI-Hv50k-2016–322771		569,020,612	569,020,612
		5 H	3.044	0.32	558,775,861	JHI-Hv50k-2016–319913		558,775,861	558,775,861
2	PL	5 H	3.276	0.29	558,776,694	JHI-Hv50k-2016–319917	3	558,776,694	558,776,694
		6 H	3.183	0.36	502,535,844	SCRI_RS_143367		502,535,844	502,535,844
3	HD	5 H	3.005	0.35	535,622 854	JHI-Hv50k-2016–313680		535,445,094	535,800,612
		5 H	3.273	0.42	535,605,717	JHI-Hv50k-2016–313622	2	535,605,717	535,605,717
4		2 H	3.002	0.52	1,766,828	JHI-Hv50k-2016–59801		1,766,828	1,766,828
		2 H	3.173	0.24	1,766,806	JHI-Hv50k-2016–59799		1,766,806	1,766,806
	FD	5 H	3.02	0.34	40,248,614	JHI-Hv50k-2016–289082		40,248,614	40,248,614
		7 H	3.105	0.32	10,251,409	JHI-Hv50k-2016–445639	7	10,251,409	10,251,409
		6 H	3.251	0.43	506195397	JHI-Hv50k-2016-411901		506,195,397	506,195,397
		2 H	3.173	0.24	1766806	JHI-Hv50k-2016-59799		1,766,806	1,766,806
		2 H	3.286	0.56	592937557	BOPA2_12_11096		592,937,557	592,937,557
		2 H	3.245	0.45	635741374	JHI-Hv50k-2016-104062		635,741,374	635,741,374
		2 H	3.387	0.36	635860279	JHI-Hv50k-2016-104089		635,860,279	635,860,279
5	MD	2 H	3.525	0.32	713474034	SCRI_RS_168629		713,474,034	713,474,034
		2 H	3.543	0.32	635741183	JHI-Hv50k-2016-104059		635,741,183	635,741,183
		2 H	3.971	0.29	636434218	JHI-Hv50k-2016-104181		636,434,218	636,434,218
		2 H	4.73	0.35	635867081	JHI-Hv50k-2016-104100		635,867,081	635,867,081
		2 H	4.248	0.24	635,549,107	JHI-Hv50k-2016-104031		635,549,107	635,549,107
		3 H	3.268	0.29	516634958	JHI-Hv50k-2016-186397		516,634,958	516,634,958
		7 H	4.404	0.45	32315380	SCRI_RS_163976	13	32,315,380	32,315,380
		7 H	3.169	0.27	583114161	JHI-Hv50k-2016-492462		583,114,161	583,114,161
		7 H	3.534	0.35	583586119	BOPA2_12_11044		583,586,119	583,586,119
		7 H	3.634	0.34	583498560	JHI-Hv50k-2016-492506		583,498,560	583,498,560
		7 H	4.269	0.43	575489630	JHI-Hv50k-2016-491866		575,489,630	575,489,630
		7 H	4.404	0.35	32315380	SCRI_RS_163976		32,315,380	32,315,380
6	GS	2H	4.268	0.39	655,112,436	BOPA1_ConsensusGBS0705-1		654,934,676	655,290,196
		1H	3.811	0.46	471,473,363	BOPA2_12_30505		471,295,603	471,651,123
		2H	4.251	0.25	591372161	JHI-Hv50k-2016-100080		591,372,161	591372161
		4H	4.276	0.42	3571618	JHI-Hv50k-2016-227520		3571618	3571618
		4H	4.151	0.52	3,156,687	JHI-Hv50k-2016-227449		3156687	3156687
		4H	4.307	0.54	600036196	JHI-Hv50k-2016-261240		600036196	600036196
		6H	4.013	0.24	545,283,678	SCRI_RS_160175		545283678	545283678
		1H	3.299	0.32	12 040013	JHI-Hv50k-2016-11522		12040013	12040013
		1H	3.091	0.43	11,665,410	JHI-Hv50k-2016-11234		11665410	11665410
		1H	3.04	0.24	12 040 644	JHI-Hv50k-2016-11528		12040644	12040644
		1H	3.046	0.56	346,196,655	SCRI_RS_7197		346196655	346196655
		1H	3.117	0.45	523970914	JHI-Hv50k-2016-46401	23	523970914	523970914
		1H	3.391	0.36	465,739,489	JHI-Hv50k-2016-36101		465739489	465739489
		2H	3.432	0.32	50345161	JHI-Hv50k-2016-77945		50345161	50345161
		2H	3.333	0.42	697,166,859	JHI-Hv50k-2016-116958		697166859	697166859
		2H	3.268	0.29	739 812 941	JHI-Hv50k-2016-134459		739812941	739812941
		2H	3.22	0.25	696,900,444	JHI-Hv50k-2016-116867		696900444	696900444
		3H	3.238	0.24	629326397	JHI-Hv50k-2016-204485		629326397	629326397
		4H	3.388	0.29	3,176,728	JHI-Hv50k-2016-227496		3176728	3176728
		5H	3.325	0.45	552961	JHI-Hv50k-2016-277134		552961	552961
		6H	3.448	0.27	575,415,722	JHI-Hv50k-2016-429157		575415722	575415722
		6H	3.243	0.35	368968887	JHI-Hv50k-2016-399571		368968887	368968887
		7H	3.348	0.34	7,071,811	JHI-Hv50k-2016-442900		7071811	7071811
7	FAN	1H	5.552	0.53	538, 390, 721	JHI-Hv50k-2016-51412		538390721	538390721
		1H	5.095	0.35	538,410,781	JHI-Hv50k-2016-51422		538410781	538410781
		1H	5.211	0.42	538505042	JHI-Hv50k-2016-51570	5	538505042	538505042
		1H	4.622	0.29	538,584,087	JHI-Hv50k-2016-51636		538584087	538584087
		1H	4.537	0.55	538503670	JHI-Hv50k-2016-51558		538503670	538503670
8	SN	7H	5.459	0.27	1,262,584	JHI-Hv50k-2016-436545		1262584	1262584
		4H	5.326	0.35	613812831	JHI-Hv50k-2016-263737		613812831	613812831
		5H	4.061	0.34	494,426,342	JHI-Hv50k-2016-308766	7	494426342	494426342
		5H	4.132	0.43	494426547	JHI-Hv50k-2016-308767		494426547	494426547
		1H	3.555	0.25	321,396,626	BOPA2_12_10300		321396626	321396626
		4H	3.511	0.37	565364643	JHI-Hv50k-2016-255317		565364643	565364643
		2H	3.214	0.45	80,834,384	JHI-Hv50k-2016-82292		80834384	80834384

MAF Minor Allele Frequency, NMTA Number of marker trait associations, SP Linkage disequilibrium decay starts Position, LDDE Linkage disequilibrium decay End Position, bp base Paire, GS Grain Size, FAN Free Amino Nitrogen, SN Soluble Nitrogen and MC Moisture Content

Discussion

The current study integrated phenotypic evaluation and GWAS to uncover genetic determinants associated with malt quality pre-indicative agronomic and nutritional traits in Ethiopian barley germplasm, leveraging a diverse panel characterized by qualitative (glume colour and row type) and quantitative traits (HD, FD, MD, PH, PL, GS, FAN, SN and MC). By combining phenotype data collected across four experimental sites with genomic SNP markers, the research aimed to identify reliable phenotypic predictors and genomic regions linked to malt quality, providing valuable insights for both the brewing industry and barley breeding programs.

Genetic admixture and breeding implications

The pairwise genetic admixture and population structure results across ten Ethiopian barley growing regions revealed significant differentiation. Regions like Gamo-Gofa (3.77) and Arsi-Bale (3.19) showed high divergence, indicating historical isolation. At the same time, Hararghe, HARC, and Gojam had low (0.30, 0.25 and 1.47) oppositely, suggesting greater connectivity, and the breeding implication is that highly divergent regions may possess unique alleles useful for introducing novel traits. In contrast, interconnected regions are valuable for developing broadly adapted barley varieties. According to Fekadu Gadissa et al. [88]., gene flow is weakly associated with geographic regions, which contrasts with the findings of this study. Abebe T.D. et al. [89]. reported that geographic regions and altitude contribute to the genetic variation observed in the germplasm, findings in line with this study. Abebe and Asmund [88] reported that geography and altitude make no significant contribution to barley agronomic traits, which slightly contradicts the findings of this study. Geographic and altitudinal variations play a crucial role in shaping the genetic diversity and gene flow patterns of barley. Environmental gradients associated with altitude, such as temperature, precipitation, and soil composition exert strong selective pressures that promote local adaptation and contribute to increased genetic differentiation among populations. Geographic isolation can further intensify this effect by restricting gene flow between regions, leading to the accumulation of unique alleles within distinct populations. These factors, together, contribute significantly to the genetic structuring observed in barley across diverse ecological zones [90, 91]. As described in Supplementary File 8, Estimates of narrow-sense heritability in the strict sense represent the proportion of phenotypic variation explained by additive genetic effects, which are the main factors in the inheritance of the trait and the response to selection. When these estimates are high, it indicates that additive genetic variance contributes substantially to the observed variation between individuals, implying that traits are strongly influenced by genes passed from parents to offspring, rather than by dominance, epistasis, or environmental effects. This is the case when the population has sufficient genetic diversity, the traits are subject to strong additive control, and the environmental noise is relatively low so that the genetic signal is clearly expressed in the phenotype [56, 92–94].

The variable genetic admixture patterns show that Pop-1 was genetically uniform, Pop-2 moderately admixed, Pop-3 mostly turquoise with minor inputs, and Pop-4 highly admixed with all clusters, reflecting extensive gene flow and the breeding implication is that diverse populations like Pop-4 are ideal for enhancing genetic variation in breeding, while uniform ones like Pop-1 can help stabilize desired traits. Earlier studies support these findings. Tiegist Dejene Abebe et al. [90] reported that 63.39% of the variation was genetic and 36.71% environmental. Alemayehu Zewdu et al. [91]. found geographic regions accounted for 39–91% of admixture and 2.72% of gene flow, while Seyyed A. Mohammadi et al. [95] reported a higher gene flow (26.89%). Breeding implies that these estimates highlight the importance of considering both genetic and environmental factors in selecting stable, high-performing genotypes. Studies by Pawan Kumar et al. [96] and Allo A. Dido et al. [97]. showed that most variation occurs among populations (78–98%), with lower within-individual diversity. The Arsi zone exhibited high gene flow, comparable to the 0.59–0.99 range observed in this study, as reported by Mihret Yirgu et al. [98]. They reported much lower gene flow (0.017–0.019). For barley breeding improvement programs, the implication is that Arsi and similarly admixed regions can serve as hotspots for accessing diverse alleles, whereas isolated populations help conserve localized adaptations. Overall, the genetic structure and admixture patterns reveal differences in evolution, ecology, and breeding practices among barley populations across Ethiopia. These insights guide parental selection, hybrid development, and targeted conservation, ultimately improving efficiency in barley breeding programs [91].

GWAS and the breeding implications

The correlations among agronomic traits in malt barley and their environmental influences are crucial for breeding programs. The negative correlation between FD and MD. In contrast, an inverse correlation between FAN and SN, which means increasing FAN as well as increasing MD suggests that early flowering leads to early maturity. In contrast, an inverse correlation between FAN and SN indicates that increasing FAN increases SN, as reported by Jerzy H. Czembor et al. [9]. This indicates that early-flowering varieties tend to be shorter. Additionally, HD, PH, and PL show a positive relationship, aligning with findings from Moustafa, Ehab S.A [99]., emphasizing the importance of plant architecture in breeding. A strong correlation between PH and peduncle length further supports this, as shorter plants may have reduced peduncle lengths, thereby impacting lodging resistance. However, the significant variation in HD, FD, MD, and PH across different experimental sites underscores the strong environmental influence on these traits, as also highlighted by Moustafa, Ehab S.A [99]., Chengsong Zhu et al. [65], and Heena Rani Rachana D. Bhardwaj [41] given that agronomic management and environmental factors significantly affect malt barley yield and quality [100]. Multi-environment trials are crucial for identifying stable accessions. Breeding programs must consider both genetic correlations and environmental interactions to optimize yield, quality, and adaptability, ensuring the development of resilient barley varieties suited for diverse growing conditions [65].

In this study, a GWAS was conducted with a genetically diverse panel of 260 mapped individuals, resulting in the identification of 65 markers with strong associations between markers and traits. The P-values of these markers ranged from 3.0 to 5.8, with significant associations observed between 4 and 5.8. Different levels of significance were found for different traits. For PH, two SNPs on chromosomes 2 and 5, namely JHI-Hv50k-2016–114427 and JHI-Hv50k-2016–114347, showed strong associations. For PL, three SNP markers were identified on chromosomes 5 and 6, including JHI-Hv50k-2016–319913, SCRI_RS_143367 and JHI-Hv50k-2016–319917. In relation to HD, two SNP markers on chromosome 5, JHI-Hv50k-2016–313680 and JHI-Hv50k-2016–313622, showed significant associations. The FD trait was associated with seven SNP markers on chromosomes 2, 5, 6 and 7, with BOPA2_12_11096 and JHI-Hv50k-2016–411901 showing particularly strong associations. Notably, FD was the second most frequently associated trait with multiple marker-trait associations. Trait MD showed associations with 13 markers on chromosomes 2, 3 and 7. Among them, SCRI_RS_163976, JHI-Hv50k-2016–491866, JHI-Hv50k-2016–104181 and JHI-Hv50k-2016–104031 had strong marker-trait associations. The GS trait had the highest MAT associations, namely 23 significant SNPs located on chromosomes 1, 2, 4, 6 and 7. Among the significant SNPs, the most are BOPA1_ConsensusGBS0705-1, JHI-Hv50k-2016–100080, JHI-Hv50k-2016–227520, JHI-Hv50k-2016–227449, SCRI_RS_160175. Teklemariam et al. [70] identified 23 MTAs under control conditions using the BLINK model, primarily located on chromosome 2, including at the Holeta site, which was also utilized. In contrast, we identified 98 MTAs at Holeta, distributed across all chromosomes. While they found MTAs for plant height under control conditions, our study also detected 4 stable SNPs for this trait across multiple locations. Additionally, their drought-stress MTAs for plant height [17] were from Melkassa and Dera, sites different from ours (Holeta, Welkite, Debre Markos, and Bekoji), which were used to assess malt quality traits under non-stress conditions. The other thing is that Teklemariam et al. [6]. didn’t incorporate the 4 major traits in this manuscript, GS, FAN, SN, and MC, which are considered the predictive traits for malt quality study and have significant SNPs for those traits. These differences are expected due to distinct research objectives and environmental conditions. Since the traits studied are quantitative, they are influenced by both genetic and environmental factors [88, 101, 102], and their expression varies across environments [103–105]. This explains the variation in MTAs and trait responses between the two studies and highlights the importance of context-specific evaluation in barley breeding [56, 106].

Several previous studies have reported marker-trait associations for these traits, supporting these findings. For instance, Jerzy H. Czembor et al. [9] identified SNP markers associated with agronomic traits, finding five SNPs for HD on chromosome 2, five for MD, and eight for PH. These results align with the findings of this study, except for the marker located on chromosome 5. The chromosome identified for HD in this study is entirely consistent with the findings of Czembor et al. [9]. Similarly, the findings for MD share partial similarity, with additional associations identified on chromosomes 3 and 7. Likewise, Rajiv Sharma et al. [107] reported SNP markers on chromosomes 6 and 7 associated with FD, which supports the associations found in this study. In contrast, Laura Paire et al. [57]. identified 1,635 SNP markers associated with agronomic traits using 2,024 different models, a finding that differs significantly from this study due to the larger number of markers detected and the use of multiple association models that yielded varying results. Additionally, Jeannette Lex et al. [108] reported 52 SNP markers associated with agronomic traits, while Xin Hu et al. [109] and Mitra Jabbari et al. [110] identified 913 and 167 SNP markers, respectively. Daba et al. [55] identified 106 significant SNPs, whereas in this study we identified 65 significant SNPs, especially for the trait FAN, with significant SNPs on chromosomes similar to those in the current study; however, the remaining parameters used differ from those in the current study. The differences in the findings between these studies and the present one can be attributed to variations in the GWAS models used and, in the number, and diversity of mapping populations. Overall, these findings contribute to a growing body of knowledge on the genetic basis of key malt quality pre-indicative agronomic traits in barley, providing valuable insights for breeding programs aimed at improving crop performance.

Conclusion

This study integrates phenotypic and genomic analyses to identify key agronomic traits influencing malt quality in Ethiopian barley. By evaluating qualitative traits (glume colour, row type) and quantitative traits (HD, FD, MD, PH, PL, GS, FAN, SN and MC), alongside correlation, PCA, clustering, Genetic admixture and gen flow nature of the mapping panel, LD decay, and GWAS, significant phenotypic variability and marker-trait associations were revealed. GWAS identified both novel and known SNPs associated with malt-related traits, underscoring the untapped genetic potential of Ethiopian germplasm for barley improvement. Uneven SNP distribution and LD decay patterns across chromosomes offer important clues for fine mapping and targeted selection. Genetic admixture and population structure analyses further revealed substantial diversity and gene flow among regional populations, with Arsi and Gamo-Gofa showing distinct divergence. These findings not only inform strategic parental selection but also support Marker-assisted and genomic selection approaches. The combined insights offer a pathway for developing high-quality malting barley varieties tailored to diverse environments and industry needs. Moving forward, validation of key traits and SNPs under industrial processing conditions will be essential to ensure practical relevance and accelerate precision breeding for improved malt quality and market competitiveness in Ethiopia

Abbreviations

BC

Black colour

FD

Flowering Day

GAPIT

Genome Association and Prediction Integrated Tool

GL

Glume colour

GWAS

Genome Wide Association Study

HD

Heading Day

LD

Linkage disequilibrium

MD

Maturity Day

NJ

Neighbor-Joining

PC

Purple colour

PH

Plant Height

PL

Peduncle Length

RN

Row Number

SC

Six Row

SC

Straw colour

SNP

Single Nucleotide Polymorphism

TR

Two Row

Two Row

Yellow colour

Authors’ contributions

B.B. conceived and designed the paper, analyzed the experimental data, and wrote the manuscript. A.T., W.F., G.H., S.K., S.S. executed and revised the manuscript and A.A.W. executed and revised the manuscript and supervised and administrated the project. The final manuscript was approved by all authors.

Funding

This study was conducted with institutional support from the Ethiopian Institute of Agricultural Research, the National Agricultural Biotechnology Research Centre, and Addis Ababa Science and Technology University, although no specific research funding was allocated.

Data availability

All the genotypic data are available on paper and its supporting information files. The SNP data is available online at https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0260422.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.