Genomic selection and association analyses for improving semen traits in Thai native roosters Authors: Jiraporn Juiputta, Veeraya Tuntiyasawasdikul, Vibuntita Chankitisakul, Wuttigrai Boonkum Published: 2026 ABSTRACT AbstractSemen quality is a major determinant of reproductive efficiency in poultry production systems. In Thai native chickens, the relatively low male-to-female mating ratio increases the reproductive importance of individual sires, making semen quality traits critical for maintaining flock fertility. Thai native chickens represent an important genetic resource characterized by high adaptability to tropical environments and unique genetic diversity, yet remain relatively under-studied compared with commercial lines. Therefore, accurate estimation of genetic parameters and a clear understanding of the genetic architecture underlying semen traits are required to support effective breeding programs. However, despite their importance, semen quality traits in Thai native chickens remain poorly characterized at the genomic level, and the effectiveness of genomic evaluation approaches for improving these traits has not been fully explored. This study estimated genetic parameters, characterized the genetic architecture of semen traits, and evaluated the impact of genomic information on prediction accuracy in Thai native chickens. We evaluated the predictive performance of sperm quality traits using three genetic evaluation models: pedigree-based best linear unbiased prediction (PBLUP), single-step genomic BLUP (ssGBLUP), and weighted ssGBLUP (WssGBLUP). Estimates of heritability for mass movement, semen volume, and sperm concentration were low (0.149, 0.165, and 0.060, respectively), indicating limited additive genetic variation for these traits. Incorporation of genomic information improved prediction accuracy, with ssGBLUP and WssGBLUP increasing accuracy by 5.88% and 7.49%, respectively, compared with PBLUP. A strong positive genetic correlation (0.977) was observed between mass movement and sperm concentration, suggesting that selection for improved mass movement is likely to result in a favorable correlated genetic response in sperm concentration. In addition, WssGWAS identified genomic regions on chromosomes 6, 11, 19, 21, and 22 that harbored 14 candidate genes associated with sperm structure and metabolic processes. These findings suggest that the identified genomic regions may influence semen quality through biological pathways related to sperm membrane integrity, energy metabolism, and cellular regulation, which are essential for maintaining sperm motility and function. Our results demonstrate the utility of genomic information, particularly the WssGBLUP approach, for improving semen quality traits in Thai native chickens. These findings provide biologically relevant insights to support the development of sustainable genetic improvement programs.Keywords: Semen quality, Genomic prediction, Genetic parameters, Indigenous chicken Introduction Optimizing reproductive efficiency is essential for poultry production systems to meet increasing demand for chicks. Although both sexes contribute to reproductive performance, roosters exert a disproportionate influence due to the low male-to-female ratio used in breeding flocks (Sun et al., 2019). Consequently, the reproductive capacity of individual sires plays a critical role in determining flock fertility. Semen quality traits, including semen volume, sperm motility (mass movement), and sperm concentration, are key determinants of fertilization rate, hatchability, and embryo viability, and are commonly used as selection criteria in breeding programs (Tesfay et al., 2020; Mussa et al., 2023; Juiputta et al., 2025). However, evaluation based on individual semen traits may not fully reflect reproductive performance because these traits are biologically interrelated. Previous studies have shown that multivariate approaches improve prediction of fertility compared with single-trait evaluations (Tesfay et al., 2020; Daryatmo et al., 2024). Most semen quality traits exhibit low to moderate heritability (h² ≈ 0.05–0.30), indicating a complex and polygenic genetic architecture. Nevertheless, traits such as semen volume, mass movement, and sperm concentration often show relatively higher heritability and may respond to genetic selection (Juiputta et al., 2025). Thai native chickens, including Pradu Hang Dum and related indigenous breeds, are important genetic resources characterized by adaptation to tropical environments and high genetic diversity (Loengbudnark et al., 2023; Tanglertpaibul et al., 2024). Despite their importance, genetic improvement programs in these populations remain largely based on phenotypic selection and pedigree-based evaluation called pedigree-based best linear unbiased prediction (PBLUP). The effectiveness of these approaches is limited by incomplete pedigree information and reduced accuracy for low-heritability traits (Kim et al., 2023; Misztal et al., 2020). In addition, genetic and genomic information for semen quality traits in Thai native chickens remains limited. Genomic selection provides an opportunity to improve prediction accuracy by incorporating dense single-nucleotide polymorphism (SNP) data to estimate genomic breeding values (Meuwissen et al., 2001). Among genomic approaches, single-step genomic best linear unbiased prediction (ssGBLUP) integrates pedigree, phenotypic, and genomic information into a unified framework and allows evaluation of both genotyped and non-genotyped animals (Misztal et al., 2020). Although ssGBLUP has improved prediction accuracy in several livestock species, its performance for semen quality traits in indigenous chicken populations has not been well established. Moreover, standard ssGBLUP assumes equal variance across markers, which may limit its ability to capture loci with moderate or large effects. Weighted ssGBLUP (WssGBLUP) extends ssGBLUP by assigning marker-specific weights based on estimated genetic variance, which may improve prediction accuracy and the detection of biologically relevant genomic regions (Teissier et al., 2019). When combined with genome-wide association studies (GWAS), this approach can provide insight into the genetic architecture of complex traits. Although GWAS has identified candidate genes for reproductive traits in some chicken populations (Zhang et al., 2023; Zhu et al., 2025), information on semen quality traits in Thai native chickens remains limited. In particular, comparative evaluation of genomic prediction models and characterization of associated genomic regions for these traits have not been well established. This limitation may constrain the effective application of genomic selection for improving reproductive performance in indigenous chicken populations. Therefore, the objectives of this study were to (1) compare the predictive performance of PBLUP, ssGBLUP, and WssGBLUP for semen quality traits and (2) identify genomic regions and candidate genes associated with these traits using GWAS in Thai native chickens. Materials and methods Ethics statement The Institutional Animal Care and Use Committee approved the experimental procedures following the National Research Council of Thailand's Ethics of Animal Experimentation (Record no. IACUC-KKU-110/68). Data and resource population The dataset comprised 5,597 records from 395 Thai native Pradu Hang Dum grandparent males aged 40–156 weeks. Birds were housed individually in open-sided cages (45 × 50 × 60 cm) under natural photoperiod and ambient temperature conditions. All birds were provided approximately 110 g/day of a commercial breeder diet containing 90.07% dry matter (DM), 17.15% crude protein (CP), 3.35% crude fiber (CF), 3.99% ether extract (EE), and 9.75% ash, with ad libitum access to fresh water. Birds were reared from hatch to the end of the study at the Network Center for Animal Breeding and Omics Research, Faculty of Agriculture, Khon Kaen University, Thailand. Semen quality data were collected from roosters and included three traits: sperm concentration (× 109 sperm/mL), semen volume (mL/ejaculation), and mass movement (score). Mass movement was assessed on an ordinal categorical score (1–5) under light microscopy by a single trained technician to minimize evaluator bias. Phenotypic data Semen was collected using the Dorso abdominal massage method (Castillo et al., 2021) and immediately diluted with 0.1 mL of modified IGGKPh extender (Chankitisakul et al., 2022). Collections were performed by trained personnel to minimize contamination. Samples were protected from light and maintained at 22–25°C during transport to the laboratory, which occurred within 20 min of collection. Semen volume was measured directly using a 1-mL graduated syringe. Sperm motility was evaluated as mass movement by placing a drop of undiluted semen on a glass slide without a coverslip and examining wave motion under a compound microscope (100 × magnification). Mass movement was scored on a scale of 1–5. Sperm concentration was determined using a hemocytometer. Briefly, 1 µL of semen was diluted in 999 µL of 4% (wt/vol) sodium chloride solution (1:1,000 dilution). The diluted sample was loaded into the counting chamber, and sperm cells were counted under a light microscope (400 × magnification). Sperm concentration was calculated and expressed as spermatozoa per milliliter. Data structure and descriptive statistics for semen traits in Thai native chickens are presented in Table 1. Table 1Data structure and descriptive statistics of semen traits in Thai native chickens.Table 1 dummy alt textItemNumberMean ± SDContemporary Groups Hatch × generation (n)28Animal Records Number of animals with records (n)395 Number of animals with pedigree (n)826 Number of animals with genotype (n)287Traits Mass movement (score)5,5973.60 ± 1.18 Semen volume (mL/ejaculation)5,5970.32 ± 0.16 Sperm concentration (× 109 sperm/mL)5,5973.66 ± 1.46SD = Standard deviation. Data structure and descriptive statistics of semen traits in Thai native chickens. SD = Standard deviation. Genotypic data and quality control Blood samples were collected from the wing veins of 287 chickens (150 males and 137 females) by standard venipuncture. All birds remained clinically normal following sampling and throughout the experimental period. Genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany; Cat. No. 69504) according to the manufacturer’s instructions. DNA concentration and purity were assessed using a NanoDrop spectrophotometer, and samples with a concentration ≥40 ng/µL were selected for genotyping with the Axiom Genome-Wide Chicken 580 K SNP Array. Genotype quality control was performed using the BLUPF90+ software family, version 2.56 (Misztal et al., 2024). SNPs and individuals with a call rate <0.90 were excluded. Markers with a minor allele frequency (MAF) <0.05 and monomorphic SNPs were also removed. Parent–progeny conflicts were evaluated, and SNPs with >10% conflicts and individuals with >1% conflicts were discarded. After quality control, 434,173 SNPs remained for subsequent genomic selection analyses. Genetic evaluation models Genetic parameters (heritability, genetic correlations, and permanent environmental correlations) and breeding values were estimated using a multiple-trait repeatability threshold–linear model within a Bayesian framework, implemented in GIBBSF90 (BLUPF90+ software family, version 2.56). Mass movement was treated as an ordinal trait (scored 1–5) using a threshold model, whereas semen volume and sperm concentration were treated as continuous traits using a linear model. The model is defined as follows: where y is the vector of observed phenotypes (liability scale for the threshold trait); b is the vector of fixed effects, including hatch–generation set, rooster age, and body weight; a is the vector of additive genetic effects; pe is the vector of permanent environmental effects; and e is the vector of residual effects. X, Z, and W are incidence matrices relating observations to fixed, additive genetic, and permanent environmental effects, respectively. PBLUP Genetic relationships among animals were modeled using the numerator relationship matrix (A), which is derived from pedigree information. Additive genetic effects were assumed to follow a multivariate normal distribution, as follows: where σa2 denotes the additive genetic variance. ssGBLUP The ssGBLUP approach was applied to jointly incorporate pedigree and genomic information. In this method, the inverse of the pedigree-based relationship matrix in the mixed model equations is replaced by the inverse of the combined relationship matrix (H), defined as follows: where H−1 is the inverse of the combined relationship matrix, A−1 is the inverse of the pedigree-based relationship matrix, G−1 is the inverse of the genomic relationship matrix, and A22−1 is the inverse of the pedigree-based relationship matrix corresponding to genotyped animals. WssGBLUP A WssGBLUP approach with a multi-trait genetic model was used for genomic evaluation. In contrast to the traditional animal model based solely on the inverse of the numerator relationship matrix (A), the inverse of the combined relationship matrix (H) was incorporated into the mixed-model equations, as follows: where A−1 is the inverse of the pedigree-based relationship matrix for all animals included in the analysis, G−1 is the inverse of the genomic relationship matrix, and A22−1 is the inverse of the pedigree-based relationship matrix for genotyped animals only. As only incomplete pedigree information was available, inbreeding coefficients were not included in the construction of matrix A, which may have slightly affected the estimates. The genomic relationship matrix (G) and SNP weights were constructed following the method described by VanRaden (2008): where Z is the matrix of centered genotype covariates adjusted for allele frequencies, D is a diagonal matrix containing SNP-specific weights, m is the total number of SNP markers, pi is the reference allele frequency for the ith SNP, di is the weight assigned to the ith SNP, and s^i is the posterior mean SNP effect obtained from the threshold ssGBLUP analysis. All analyses were performed using Gibbs sampling with 120,000 iterations per analysis. The first 60,000 iterations were discarded as burn-in to ensure convergence of the Markov chains. Thereafter, one sample was retained every 20 iterations to reduce autocorrelation among successive samples. Posterior means of the retained samples were used to estimate variance components, estimated breeding values (EBVs), and genomic estimated breeding values (GEBVs). Convergence was evaluated by visual inspection of trace plots for key parameters. Model comparison The predictive performance of PBLUP, ssGBLUP, and WssGBLUP was evaluated using the linear regression (LR) method described by Legarra and Reverter (2018). The dataset was randomly partitioned into complete (w) and partial (p) datasets, where the partial dataset was generated by masking phenotypic records from a subset of animals. In this study, phenotypes of 75 validation animals (50% of current-generation roosters) were masked to create the partial dataset. Breeding values were estimated from both datasets, and prediction accuracy, bias, and dispersion were calculated based on comparisons between estimates obtained from the complete and partial datasets. Model performance for this validation group was assessed using the following indicators: (1)Bias: Bias was calculated to quantify the mean difference between EBVs/GEBVs obtained from the partial and whole datasets:μw,p=u^‾p−u^‾w where u^‾p and u^‾w are the mean predicted breeding values from the partial and whole datasets, respectively. Bias: Bias was calculated to quantify the mean difference between EBVs/GEBVs obtained from the partial and whole datasets:μw,p=u^‾p−u^‾w where u^‾p and u^‾w are the mean predicted breeding values from the partial and whole datasets, respectively. (2)Accuracy: Accuracy was assessed as the Pearson correlation between EBVs/GEBVs estimated from the whole and partial datasets, reflecting the model’s ability to correctly rank animals:ρw,p=cov(u^p,u^w)var(u^p)var(u^w) Accuracy: Accuracy was assessed as the Pearson correlation between EBVs/GEBVs estimated from the whole and partial datasets, reflecting the model’s ability to correctly rank animals:ρw,p=cov(u^p,u^w)var(u^p)var(u^w) (3)Dispersion: Dispersion was evaluated using the regression coefficient of u^w on u^p. An ideal value of 1.00 indicates absence of over- or under-dispersion:βw,p=cov(u^w,u^p)var(u^p) Dispersion: Dispersion was evaluated using the regression coefficient of u^w on u^p. An ideal value of 1.00 indicates absence of over- or under-dispersion:βw,p=cov(u^w,u^p)var(u^p) (4)Covariance: The covariance between predicted breeding values from the whole and partial datasets was calculated to assess their linear association:cov(u^w,u^p) Covariance: The covariance between predicted breeding values from the whole and partial datasets was calculated to assess their linear association:cov(u^w,u^p) GWAS To identify genomic regions and SNPs associated with semen traits in Thai native roosters, a GWAS was performed from the best model. SNP effects were estimated iteratively using the POSTGSF90 software package. The additive genomic breeding values (u^) were back-transformed to SNP effects (s^) based on their shared genomic variance (σu2) using the following equation: where Dis a diagonal matrix of SNP weights, Z is the incidence matrix relating SNP genotypes to individuals, and G is the genomic relationship matrix. The proportion of additive genetic variance explained by each SNP window (five consecutive SNPs) was calculated as follows: where Var(∑j=15zjs^j) represents the genetic variance explained by a given five-SNP window, σ^u2 is the total additive genetic variance, zj is the vector of gene content for the jth SNP across all individuals, and s^j is the estimated effect of the jth SNP. To identify candidate genes associated with the three semen traits, genomic regions explaining more than 0.5% of the total additive genetic variance were considered. Genomic coordinates were annotated based on the Gallus gallus reference assembly (GRCg4). For each significant SNP, a 1 Mb window (0.5 Mb upstream and 0.5 Mb downstream) was defined. Candidate genes within these regions were retrieved from the NCBI database, with emphasis on functionally characterized genes. Functional annotation and pathway enrichment analyses were performed using DAVID Bioinformatics Resources (version 2025_2). All statistical analyses and visualizations, including Manhattan plots, were conducted using R (version 4.5.2) within the RStudio environment (version 2026.01.2). Results Impact of chicken age on semen traits The relationship between age (40–156 weeks) and semen quality traits is shown in Fig. 1. Sperm concentration and mass movement progressively declined with advancing age, with sperm concentration exhibiting the steepest negative slope (b = −0.008). In contrast, semen volume remained relatively stable (mean = 0.320 mL; b ≈ 0.000). A more pronounced decline and increased variability were observed after 100 weeks of age, indicating reduced semen quality at advanced ages. These patterns suggest that age-related deterioration primarily affects functional sperm traits rather than ejaculate volume. Age-related changes in semen traits in Thai native chickens. Impact of age on breeding values for semen traits As shown in Fig. 2, breeding values (EBVs and GEBVs) for all semen traits decreased with advancing age. The strongest decline was observed for sperm concentration (b = −0.025), followed by semen volume (b = −0.013), whereas mass movement showed minimal change (b = −0.002), indicating greater genetic stability. Increased variability in breeding values after 100 weeks suggests less consistent genetic expression at older ages, which may influence selection decisions in breeding programs. Relationship between chicken age and estimated breeding values (EBVs; light blue circles) and genomic estimated breeding values (GEBVs; dark blue circles) for semen traits in Thai native chickens. Genetic parameter estimates The estimated genetic parameters for mass movement, semen volume, and sperm concentration obtained using PBLUP, ssGBLUP, and WssGBLUP are presented in Table 2. Heritability estimates were low across all traits, ranging from 0.046 to 0.143 under PBLUP and increasing to 0.060–0.165 under WssGBLUP. The inclusion of genomic information increased additive genetic variance and improved heritability estimates, particularly for semen volume. Genetic correlations were stronger under WssGBLUP, especially between semen volume and sperm concentration (0.077 to 0.287), indicating shared genetic control. In contrast, permanent environmental correlations decreased across models, suggesting improved partitioning of genetic and environmental effects. These results indicate that genomic models better capture the genetic architecture of semen traits. Table 2Comparison of variance components and genetic parameter estimates for semen traits in Thai native chickens obtained using pedigree-based best linear unbiased prediction (PBLUP), single-step genomic best linear unbiased prediction (ssGBLUP), and weighted single-step genomic best linear unbiased prediction (WssGBLUP).Table 2 dummy alt textParametersTraitsPBLUPssGBLUPWssGBLUPσa2Mass movement0.1880.1850.197Semen volume0.2570.3800.374Sperm concentration0.0920.1110.121σpe2Mass movement0.0380.0470.030Semen volume0.4640.3610.360Sperm concentration0.3130.2970.282σe2Mass movement1.0901.0911.090Semen volume1.5271.5271.527Sperm concentration1.6071.6071.607h2±SEMass movement0.143 ± 0.0320.140 ± 0.0350.149 ± 0.034Semen volume0.114 ± 0.0510.168 ± 0.0480.165 ± 0.048Sperm concentration0.046 ± 0.0230.055 ± 0.0250.060 ± 0.023rgMass movement vs. Semen volume0.2470.3070.329Mass movement vs. Sperm concentration0.8990.9790.977Semen volume vs. Sperm concentration0.0770.2640.287rpeMass movement vs. Semen volume0.3650.1940.086Mass movement vs. Sperm concentration0.8030.6360.601Semen volume vs. Sperm concentration0.6260.6490.628σa2 = additive variance; σpe2= permanent environmental variance; σe2= residual variance; h2 = heritability; SE = Standard error; rg = genetic correlation; rpe = permanent environmental correlation. Comparison of variance components and genetic parameter estimates for semen traits in Thai native chickens obtained using pedigree-based best linear unbiased prediction (PBLUP), single-step genomic best linear unbiased prediction (ssGBLUP), and weighted single-step genomic best linear unbiased prediction (WssGBLUP). σa2 = additive variance; σpe2= permanent environmental variance; σe2= residual variance; h2 = heritability; SE = Standard error; rg = genetic correlation; rpe = permanent environmental correlation. Validation performance The predictive performance of the breeding models for semen traits in Thai native chickens is presented in Table 3. Incorporating genomic information improved prediction accuracy across all traits compared with PBLUP. The greatest improvement was observed for sperm concentration (0.714 to 0.823 and 0.806 for ssGBLUP and WssGBLUP, respectively). WssGBLUP showed the highest overall accuracy (mean = 0.804) and reduced dispersion, indicating more stable predictions. Bias values were low (0.002–0.003) across all models, confirming minimal systematic deviation. These results demonstrate the advantage of genomic models for improving prediction reliability in semen traits. Table 3Validation performance and comparison of genetic evaluation models using the LR method.Table 3 dummy alt textTraitsStatisticsPBLUPssGBLUPWssGBLUPMass movementBias0.0030.0020.002Dispersion1.1951.1431.118Accuracy0.7370.7740.795Covariance0.0160.0270.033Semen volumeBias0.0010.0040.004Dispersion0.5860.8440.958Accuracy0.7940.7810.810Covariance0.0430.0580.063Sperm concentrationBias0.0020.0040.003Dispersion0.5430.9380.865Accuracy0.7140.8230.806Covariance0.0090.0280.027Overall averageBias0.0020.0030.003Dispersion0.7750.9750.980Accuracy0.7480.7920.804Covariance0.0220.0380.041 Validation performance and comparison of genetic evaluation models using the LR method. SNP effects and genomic regions As shown in Fig. 3, the Manhattan plots indicate that semen traits are influenced by multiple loci with small effects distributed across several chromosomes. Notably, regions on chromosomes 6, 11, 19, 21, and 22 explained more than 0.5% of the genetic variance and were consistently identified across traits, including mass movement, semen volume, and sperm concentration. The detection of overlapping genomic regions affecting multiple traits is consistent with the strong genetic correlations observed among these semen characteristics. Manhattan plots showing proportion of additive genetic variance explained by sliding windows of five adjacent single-nucleotide polymorphisms (SNPs) for mass movement, semen volume, and sperm concentration in Thai native chickens. Y-axis represents proportion of additive genetic variance attributed to each genomic window, and x-axis indicates chromosomal position. Horizontal line in each panel denotes the threshold for windows explaining ≥0.5% of total additive genetic variance. Identification of candidate genes and significant SNPs for semen traits GWAS identified several significant genomic windows harboring candidate genes associated with semen traits (Table 4). For semen volume, significant regions were detected on chromosomes 6 and 21, including genes such as VDAC2 and GPAT2, which are involved in mitochondrial function and lipid metabolism. For mass movement and sperm concentration, overlapping regions were identified on chromosomes 11 and 19, containing genes such as RPGRIP1L, FTO, and C1QBP. These genes are associated with ciliary function, energy regulation, and cellular processes. Overall, the identified candidate genes are functionally related to mitochondrial activity, ciliary structure, and intracellular transport, which are biologically relevant to sperm function and semen quality. Table 4Candidate genes identified within genomic windows significantly associated with semen traits in Thai native chickens.Table 4 dummy alt textChroPos (Mb)N% VarTop SNPPos (bp)TraitsGenesBiological Functions615.0–16.05090.600AX-7610123015,663,271Semen VolumeVDAC2Mitochondrial porin facilitating ion/ATP/ADP transport and membrane permeability.COMTD1A SAM-dependent methyltransferase involved in O-methylation and transferase activity.SAMD8Sphingolipid synthase producing ceramide phosphoethanolamine from ceramide.DUSP13Phosphatase that dephosphorylates signaling proteins to negatively regulate the MAPK cascade.211.0–2.014900.509AX-766669821,365,289CFAP74Structural protein involved in cell division, chromosome partitioning, and cilia organization.PRKCZKinase regulating cell polarity, cytoskeleton organization, and insulin/tight junction signaling.224.0–5.09580.574AX-766938264,316,347GPAT2Mitochondrial enzyme catalyzing the first step of triglyceride and phospholipid synthesis.FAHD2AMetal-binding enzymes are involved in oxaloacetate metabolism and secondary metabolite catabolism.114.0–5.04660.758AX-763675054,877,147Mass MovementRPGRIP1LCiliary scaffold protein is essential for cilium assembly and embryonic organ development.0.777Sperm ConcentrationFTOAn oxidative demethylase that removes methyl groups from RNA to regulate lipid storage and energy homeostasis.AKTIPRegulator of endosomal vesicle trafficking, protein ubiquitination, and apoptosis.193.0–4.07720.527AX-766178533,433,845Mass MovementC1QBPMultifunctional protein involved in complement activation, RNA splicing, and innate immunity.0.522AX-766179833,467,332Sperm ConcentrationRPAINFacilitates the nuclear import of RPA complex for DNA metabolism and repair.NUP88Nuclear pore component that controls transport of proteins and mRNA in and out of the nucleus.Chro = Chromosome; Pos (Mb) = Position in Megabase; N = Number of markers; % Var = Percentage of variance explained; Top SNP = The most significant Single-Nucleotide Polymorphism in the region; Pos (bp) = Position in base pair; Trait = Phenotypic characteristic; Genes = Genes located within or near the significant region. Candidate genes identified within genomic windows significantly associated with semen traits in Thai native chickens. Chro = Chromosome; Pos (Mb) = Position in Megabase; N = Number of markers; % Var = Percentage of variance explained; Top SNP = The most significant Single-Nucleotide Polymorphism in the region; Pos (bp) = Position in base pair; Trait = Phenotypic characteristic; Genes = Genes located within or near the significant region. Discussion Reproductive efficiency is a key determinant of productivity in poultry systems, and semen traits directly influence sire fertility and flock performance (Vitorino Carvalho et al., 2021; Daryatmo et al., 2024). In this study, age-related variation in semen characteristics was evaluated, and the performance of pedigree- and genomic-based prediction models was compared, together with genome-wide association analysis. These results provide insight into the genetic architecture of semen traits and demonstrate the practical value of genomic information for improving reproductive performance in Thai native chickens. Sperm concentration and mass movement decreased with advancing age, whereas semen volume remained relatively stable. These results are consistent with previous reports indicating reduced spermatogenic efficiency and endocrine function in aging roosters (Ayeneshet et al., 2024; Roshdy et al., 2024; Guo et al., 2026). The increased phenotypic variability observed after approximately 100 weeks suggests reduced uniformity and greater environmental sensitivity in older birds. From a breeding perspective, this may reduce the reliability of phenotypic records and compromise selection consistency. Therefore, earlier selection of sires is likely to improve record stability while reducing generation interval. Heritability estimates (0.046–0.168) indicate that semen traits are under low to moderate additive genetic control, suggesting a substantial contribution of environmental and non-additive effects (Fournier et al., 2019; Wolc et al., 2019; Li et al., 2025). Under these conditions, improving prediction accuracy is essential for achieving genetic progress (Ndung’u et al., 2022; Liang and Xuan, 2024). The distribution of SNP effects across multiple chromosomes supports a polygenic architecture, with no major loci explaining a large proportion of genetic variance. This finding is consistent with the complex biological regulation of reproductive traits and supports the use of genome-wide approaches (Liang et al., 2024). Strong positive genetic correlations between sperm concentration and mass movement indicate shared biological control and suggest that these traits are influenced by common genetic mechanisms (Ayalew et al., 2026; Kalaignazhal et al., 2026). From a selection standpoint, this relationship is favorable, as improvement in one trait is expected to result in correlated improvement in the other. The absence of antagonistic correlations further supports the application of multi-trait selection strategies (Chomchuen et al., 2022; Guo et al., 2025). Genomic-based models improved prediction accuracy compared with PBLUP, particularly for sperm concentration and semen volume, reflecting the ability of genomic information to better capture realized genetic relationships (Misztal et al., 2020). The additional weighting in WssGBLUP resulted in modest, trait-dependent gains, with limited benefit for highly polygenic traits such as mass movement (Chen et al., 2025; Zhou et al., 2025). This is consistent with evidence that weighted models are most effective when traits are influenced by loci with heterogeneous effect sizes rather than uniformly small effects (Lourenco et al., 2020). In addition, genomic models enhance robustness to pedigree errors or incomplete genealogical records, improving relationship estimation and reducing bias in breeding value prediction (Zhang et al., 2019; Liu et al., 2026; Misztal et al., 2020). This advantage is particularly relevant for indigenous populations, where pedigree information may be limited. Genome-wide association analysis identified multiple genomic regions explaining small but meaningful proportions of additive variance, confirming the polygenic nature of semen traits. The detected genomic regions on chromosomes 6, 11, 19, 21, and 22 each explained more than 0.5% of additive genetic variance, indicating that semen traits are influenced by multiple loci with small to moderate effects. Candidate genes identified in this study, including VDAC2 (Park et al., 2022), CFAP74 (Zhang et al., 2024), and DUSP13 (Biscotti et al., 2020), are involved in mitochondrial function, flagellar structure, and spermatogenic processes, supporting the biological relevance of the detected signals. Additional genes such as PRKCZ, GPAT2, and RPGRIP1L are implicated in ATP regulation, genomic integrity, and flagellar development (Liu et al., 2024). Importantly, this study provides novel genomic information for semen traits in Thai native chickens, a population for which genomic evidence remains limited. However, these associations should be interpreted with caution, as they do not represent confirmed causal variants. Although the present study demonstrates the advantages of genomic evaluation, several limitations should be considered. The sample size, while sufficient for estimating genetic parameters, may limit the precision of SNP effect estimates and reduce the power to detect loci with small effects, which is a common constraint in genome-wide association studies . In addition, the SNP density used in this study may restrict mapping resolution and reduce the ability to capture causal variants in linkage disequilibrium, particularly for highly polygenic traits where many loci contribute small effects. Validation was conducted within the same population, which may lead to inflation of prediction accuracy; therefore, independent validation in larger or external populations is required to confirm the robustness and generalizability of these results. Furthermore, although environmental effects were accounted for statistically, they were not experimentally controlled, and future studies integrating physiological, hormonal, or functional genomic data would improve biological interpretation of the detected associations. In conclusion, this study demonstrates that integrating genomic information substantially enhances the prediction of semen quality traits in Thai native chickens, providing a robust framework for more precise and efficient sire selection. The improved performance of ssGBLUP and WssGBLUP, together with strong genetic relationships among traits, supports the implementation of multi-trait genomic selection strategies. In addition, the observed age-related patterns highlight the importance of optimizing sire utilization to improve selection response and reduce generation intervals. Beyond methodological advancement, these findings have direct implications for breeding program design in indigenous poultry populations. The adoption of genomic evaluation can accelerate genetic gain while maintaining the adaptability and resilience of Thai native chickens under tropical production systems. Scaling up genotyped reference populations, integrating functional genomic information, and refining trait definitions will be critical to further enhance prediction accuracy and biological interpretation. At a broader level, incorporating genomic tools into national breeding strategies offers an opportunity to improve productivity, strengthen food security, and support the conservation of locally adapted genetic resources. Collectively, this study provides a foundation for transitioning from conventional to genomics-enabled breeding in Thai native chickens and similar low-input production systems. CRediT authorship contribution statement Jiraporn Juiputta: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Veeraya Tuntiyasawasdikul: Writing – review & editing, Methodology. Vibuntita Chankitisakul: Writing – review & editing, Writing – original draft, Methodology. Wuttigrai Boonkum: Writing – review & editing, Writing – original draft, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization. Disclosures The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper Acknowledgments This work was funded by the fundamental fund of Khon Kaen University, Thailand. We thank the Network Center for Animal Breeding and Omics Research, Khon Kaen University for providing the materials and animals. References