Identification of key candidate genes for body weight traits in male Dongfeng sika deer based on genome-wide association analysis Authors: Xinyuan Zhang, Yan Zhang, Lieping Zhao, Yao Zhao, Wen-xi Qian, Guanghui Gai, Jun Song, Peize Du, Huansheng Han, Zhen Zhang Published: 2026 ABSTRACT This study aimed to identify genetic markers for body weight in adult male Dongfeng sika deer (Cervus nippon) using a genome-wide association study (GWAS). A cohort of 266 healthy, five-year-old stags was phenotyped and genotyped via whole-genome resequencing. Population structure was assessed, and a mixed linear model (MLM) was employed for association testing, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of candidate genes. The GWAS identified twelve genome-wide significant SNP loci, which were annotated to seven candidate genes: ERC2, LOC122708028, FHIT, CASP14, UBE3A, NRXN3, and GABRA5. Functional enrichment analyses revealed these genes are significantly involved in the γ-aminobutyric acid (GABA)ergic synaptic transmission pathway and pathways related to DNA damage repair and cell cycle regulation. This study provides insights into the genetic mechanisms underlying body weight traits in adult male Dongfeng sika deer and offers a reference for subsequent functional exploration of candidate genes.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-57521-2. Introduction The sika deer is one of China’s most important domesticated economic animals. Its meat, prized for its tenderness, high protein content, low fat, and low cholesterol, offers substantial nutritional benefits and has become a preferred dietary ingredient1. After the domestication and management of sika deer were implemented in China in 20202, the domestic consumer demand for venison has continued to rise, demonstrating broad development prospects. Consequently, the breeding of meat-type sika deer has been listed as a key priority in the industry’s development. In meat deer breeding, body weight serves as a core indicator for evaluating growth and development. The stable expression of its genetic traits and the continuous inheritance of genetic potential provide the fundamental guarantee for the scientific conduct of breeding work and the achievement of breed improvement goals3. Accurate evaluation and targeted improvement of growth and development traits are the core tasks in the breeding of meat-type sika deer, and body measurements and body weight, as intuitive and key phenotypic indicators, are of particularly prominent importance. Body measurements serve as quantitative representations of an animal’s physique, accurately reflecting both overall conformation and localized structural characteristics4. Among these, body weight is closely associated with production efficiency, health status, disease resistance, and environmental adaptability. In male Dongfeng sika deer, five years of age represents the typical stage of physical maturity, at which body weight is stably expressed, minimally affected by environmental fluctuations, and highly representative of adult growth potential. It provides a fundamental criterion for evaluating growth progress, management practices, and selection efficiency in breeding programs5. Generally, larger individuals exhibit robust growth, greater genetic stability, and higher antler yields6, leading to improved economic returns and better adaptability to environmental stressors and diseases. Given the multifaceted importance of body weight to both economic performance and population health in sika deer farming, targeted regulation of genes influencing this trait through genetic improvement holds great significance for sustainable industry development. Such molecular breeding strategies enhance herd productivity and genetic quality from the foundational level, ensuring long-term efficiency and stability in deer production systems. Genome-Wide Association Study (GWAS) is a robust genomic tool for dissecting the genetic architecture of complex traits by identifying statistical associations between single-nucleotide polymorphisms (SNPs) and phenotypic variation at the genome-wide level. Leveraging the principle of linkage disequilibrium (LD), the non-random association of alleles at different loci, GWAS enables the indirect capture of potential functional genetic variants that underpin trait regulation. Accordingly, this study aims to conduct a systematic investigation of body weight traits in Dongfeng sika deer stags using GWAS technology. By comprehensively scanning the entire genome for SNPs, we identified genetic loci and potential candidate genes significantly associated with body weight. This research provides new molecular markers and theoretical foundations for marker-assisted selection, genetic breeding, and scientific husbandry of Dongfeng sika deer. Materials and methods Experimental animals The experimental animal population was obtained from Dongfeng Sika Deer Breeding Co., Ltd. in Dongfeng County, Jilin Province, in 2024, totaling 266 individuals. All animals are healthy adult male sika deer, 5 years old, raised under consistent feeding and management conditions. Their body weight and body size data were measured in accordance with the methods specified in the Technical Specifications for Measuring Production Performance of Velvet Deer (NY/T 1179—2006)7. Before blood sampling, each animal was anesthetized via intramuscular injection of xylazine hydrochloride at a dose of 0.2 mg/kg body weight. All animal experiments involved in this study were conducted in strict accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, 8th ed.) and the Technical Specifications for Ethical Review of Laboratory Animal Welfare in China (GB/T 35892 − 2018), which are publicly available animal care and use standards. The animal usage protocol was reviewed and approved by the Ethics and Welfare Committee of Heilongjiang Bayi Agricultural University, with the approval number: DWKJXY2022086 and the approval date: August 2022. All operations, including animal feeding, sample collection, and related experimental procedures, were performed to minimize animal suffering and ensure animal welfare. Genomic DNA genotyping Blood samples (5 mL per individual) were collected aseptically from the jugular vein of each male sika deer in the morning before feeding. The collected samples were immediately preserved on dry ice and transported to Tianjin Jizhi Gene Technology Co., Ltd. for genomic DNA extraction, testing, and downstream analysis. DNA integrity was verified through 1.5% agarose gel electrophoresis, while concentration and purity were first measured using a NanoDrop 2000 spectrophotometer. Precise quantification was subsequently performed using a Qubit fluorometer. Samples with an OD260/OD280 ratio of 1.8–2.0 and a DNA concentration exceeding 1.5 µg/µL were selected for sequencing library preparation. Genetic data quality control SNP genotyping was performed on the DNA samples. After passing quality inspection, the DNA was randomly sheared into 350 bp fragments using a Covaris instrument. Library construction steps, including end repair, polyA tailing, adapter ligation, and PCR amplification, were completed with the Illumina TruSeq Library Construction Kit. After preliminary quantification with Qubit 2.0, verification of expected insert size via Agilent assay, and confirmation of effective concentration > 2 nM by qPCR, paired-end sequencing (PE150) was conducted on the Illumina T7 sequencing platform. The average sequencing depth per sample was ≥ 10.82×, with a genome coverage rate ≥ 97.56% (Coverage 1× ≥ 97.56%). Quality control of the raw reads was performed using fastp v1.0.1, with the following filtering criteria: trimming adapter sequences, removing paired-end reads with N content > 5%, and excluding reads in which the proportion of low-quality bases (Q ≤ 15) exceeded 40%. Finally, clean reads with an effective data rate ≥ 98.82% and Q30 ≥ 91.03% were obtained for subsequent analysis8. The genomic coordinates and gene annotations in this study are based on the caribou (Cervus elaphus) reference genome assembly mCerEla1.1 (NCBI RefSeq accession number: GCF_910594005.1) and its Annotation Release 100. Clean reads were aligned to this reference genome using Sentieon v202308.03 (which applies the same mathematical model as GATK versions 3.3–4.1) with the parameters “bwa mem -k 32 -M -R“9. After sorting and deduplication using Samtools v1.18, SNP calling was performed with the Sentieon variant detection module. The filtering criteria were as follows: Individual genotypes were retained only if the Genotype Quality (GQ) score was > 20 and supported by more than 5 sequencing reads, to eliminate low-confidence genotypes.Only biallelic SNPs were retained and multi-allelic loci were removed, as multi-allelic sites tend to increase genotyping errors and complicate association testing.SNPs with a missing rate > 20% were excluded to ensure at least 80% of individuals had valid genotypes, reducing bias caused by insufficient data.Loci with Minor Allele Frequency (MAF) < 0.05 were filtered out to avoid spurious associations from rare variants with low statistical power.SNPs with a sequencing depth ≥ 2 and quality score ≥ 20 were retained to ensure reliable variant detection and minimize false-positive calls. Individual genotypes were retained only if the Genotype Quality (GQ) score was > 20 and supported by more than 5 sequencing reads, to eliminate low-confidence genotypes. Individual genotypes were retained only if the Genotype Quality (GQ) score was > 20 and supported by more than 5 sequencing reads, to eliminate low-confidence genotypes. Only biallelic SNPs were retained and multi-allelic loci were removed, as multi-allelic sites tend to increase genotyping errors and complicate association testing. Only biallelic SNPs were retained and multi-allelic loci were removed, as multi-allelic sites tend to increase genotyping errors and complicate association testing. SNPs with a missing rate > 20% were excluded to ensure at least 80% of individuals had valid genotypes, reducing bias caused by insufficient data. SNPs with a missing rate > 20% were excluded to ensure at least 80% of individuals had valid genotypes, reducing bias caused by insufficient data. Loci with Minor Allele Frequency (MAF) < 0.05 were filtered out to avoid spurious associations from rare variants with low statistical power. Loci with Minor Allele Frequency (MAF) < 0.05 were filtered out to avoid spurious associations from rare variants with low statistical power. SNPs with a sequencing depth ≥ 2 and quality score ≥ 20 were retained to ensure reliable variant detection and minimize false-positive calls. SNPs with a sequencing depth ≥ 2 and quality score ≥ 20 were retained to ensure reliable variant detection and minimize false-positive calls. Linkage disequilibrium (LD) pruning was performed using PLINK v1.90 with a window size of 50 SNPs, a step size of 5 SNPs, and an r² threshold of 0.2, to remove redundant linked markers and obtain independent SNP markers. Functional annotation of the identified SNPs was carried out using ANNOVAR v20231023. Three categories of annotation were included: gene-based annotation (e.g., exon region, intron region), chromosome interval-based annotation, and variation data-based annotation. This step clarified the genomic location, mutation type, and corresponding gene information of the variant loci10. Population structure analysis Population stratification refers to the presence of subpopulations in which individuals exhibit closer genetic relatedness than the population average. Variation in allele frequencies across subpopulations may introduce false positives in association analysis. Therefore, only independent markers were used for population stratification analysis. After filtering out linked loci, the remaining SNPs were subjected to population stratification assessment. Population phylogenetic tree analysis A phylogenetic tree, a branching diagram depicting evolutionary relationships among populations, was constructed to illustrate divergence patterns and genetic relatedness between individuals. Genetic affinities and divergences were inferred from shared or distinct genotypic characteristics, with the distance matrix calculation and phylogenetic tree construction performed using TreeBeST v1.9.211. Principal component analysis Principal Component Analysis (PCA) was applied to genotype data during preprocessing to infer population genetic structure11,12, using GCTA v1.9.4.113. Raw SNP genotype data were standardized to eliminate bias from locus-specific allele frequency differences, following the formula:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$xij^{\prime} = \frac{{xij - \bar{x}}}{{Sj}}$$\end{document} where xij denotes the genotype value of the ith sample at the jth SNP locus, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\stackrel{\mathrm{-}}{\mathrm{x}}$$\end{document}j is the mean genotype value at the jth locus, and Sj represents the standard deviation at the jth locus. The first two principal components (PCs) were then extracted and used to generate a two-dimensional scatter plot for visualizing the genetic clustering patterns among all samples. Analysis of population genetic structure Population genetic structure was inferred to characterize non-random genetic variation distribution within the study cohort, using Admixture v1.3.014. The number of putative genetic clusters (K) was tested from 2 to 6 (in the absence of prior subgroup information), with cross-validation (CV) performed for each K value. The optimal K was defined as the value with the lowest cross-validation error, representing the best-fit population stratification model15. Genome-wide association study (GWAS) For the association analysis of body weight trait in this study, GEMMA (v0.98.5)(http://www.xzlab.org/software.html) was employed, with statistical modeling performed using a Mixed-Linear Model (MLM)16,17. In the model, the genetic relatedness among individuals was set as a random effect. Age was not included as a fixed effect because all experimental deer were uniformly 5 years old with negligible age variation. The first three principal components (PC1–PC3) were included as fixed effects to correct for population stratification and reduce the false positive rate. No additional batch effect was included, as all animals originated from the same breeding center under consistent management. Detailed information on animal grouping and fixed-effect selection is provided in Supplementary File 1. The MLM model is as follows:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$y = Xa + Z\beta + W\mu + e$$\end{document} Here: y represents the phenotype value of body weight (after normality test and removal of extreme values, Mean ± 3SD is the criterion for determining outliers); X denotes the fixed effect matrix including the first three principal components (PC1–PC3) from Principal Component Analysis (PCA), which were used to correct for population stratification; Z is the SNP genotype matrix; W is the random effect matrix; α is the fixed effect vector, β is the SNP effect vector, µ is the random effect vector, and e is the residual vector (e ~ N (0, Iσ²e)). Functional annotation and enrichment analysis of candidate genes This study identified SNP loci significantly associated with body weight traits using GWAS, and the genome-wide significance threshold was determined via Bonferroni correction to rigorously control for multiple testing18. Bonferroni-corrected significance level follows the following formula:a = 0.01/N In this correction, N denotes the total number of high-quality filtered SNPs included in the association analysis. Given approximately 28 million qualified SNPs and a nominal α threshold of 0.01, the theoretical Bonferroni-corrected significance level was calculated as 3.6 × 10⁻¹⁰. To further reduce false-positive signals arising from population stratification and residual confounding, we adopted a more stringent threshold of P < 1 × 10⁻¹⁰ (-log₁₀(P) > 10) to ensure the robustness and reliability of detected associations. Genes within the candidate genomic regions surrounding significant SNPs were extracted as potential candidate genes, and their functional annotations were obtained using NCBI and Ensembl databases. The DAVID database (https://david.ncifcrf.gov/) was then employed to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses, enabling systematic interpretation of the biological functions and pathways underlying body weight variation19. KEGG pathway maps (map04080, map05033) were used under copyright permission from Kanehisa Laboratories. Data statistics Body weight trait data were collated using Excel and statistically analyzed with SPSS 26.0. Results Phenotypic data analysis of various indicators of body weight and body measurements The phenotypic characteristics of the experimental population are summarized in Table 1. Body weight of sika deer ranged from 105.50 kg to 173.50 kg, with a mean value of 136.53 ± 15.60 kg. Correspondingly, the body slant length, body height, and chest circumference exhibited moderate levels of phenotypic variation, with the mean values reaching 110.63 ± 5.73 cm, 104.24 ± 4.26 cm, and 63.10 ± 3.55 cm, respectively. It should be noted that only body weight was used as the phenotype in the GWAS; body measurement traits are presented solely for descriptive purposes. These data indicate a wide range of phenotypic diversity in the studied population, providing a reliable basis for subsequent genome-wide association analysis.Table 1Phenotypic data of body weight and body measurements indicators.ProjectWeight (kg)Body slant length (cm)Body height (cm)Circumference of chest (cm)Chest depth (cm)Head length (cm)Forehead width (cm)Girth (cm)Tail length (cm)Angle handle distance (cm)Minimum value105.5097.5095.5058.0042.5022.0012.5010.007.507.50Maximum value173.50122.00115.5072.5055.5036.0016.0013.5019.5013.50Mean ± Standard Deviation136.53 ± 15.60110.63 ± 5.73104.24 ± 4.2663.10 ± 3.5548.32 ± 3.2632.36 ± 2.8314.45 ± 0.8411.82 ± 0.5916.56 ± 1.9210.19 ± 1.13Body weight was the primary trait analyzed in the GWAS. Body measurement traits (body slant length, body height, chest circumference, etc.) are shown here only to describe the general population structure and are not included in the association analysis. Table 1Phenotypic data of body weight and body measurements indicators.ProjectWeight (kg)Body slant length (cm)Body height (cm)Circumference of chest (cm)Chest depth (cm)Head length (cm)Forehead width (cm)Girth (cm)Tail length (cm)Angle handle distance (cm)Minimum value105.5097.5095.5058.0042.5022.0012.5010.007.507.50Maximum value173.50122.00115.5072.5055.5036.0016.0013.5019.5013.50Mean ± Standard Deviation136.53 ± 15.60110.63 ± 5.73104.24 ± 4.2663.10 ± 3.5548.32 ± 3.2632.36 ± 2.8314.45 ± 0.8411.82 ± 0.5916.56 ± 1.9210.19 ± 1.13Body weight was the primary trait analyzed in the GWAS. Body measurement traits (body slant length, body height, chest circumference, etc.) are shown here only to describe the general population structure and are not included in the association analysis. Phenotypic data of body weight and body measurements indicators. Body weight was the primary trait analyzed in the GWAS. Body measurement traits (body slant length, body height, chest circumference, etc.) are shown here only to describe the general population structure and are not included in the association analysis. Results of data quality control and SNP filtering The original whole-genome resequencing detected a total of 27,127,458 SNP loci. After population-level screening, 2,798,453 high-quality SNPs were retained. LD pruning resulted in the final acquisition of 116,024 independent SNP markers for population stratification analysis. The specific screening process is shown in Table 2.Table 2Summary of SNP filtering and quality control steps.Screening stepsThe remaining number of SNPsSelection criteriaInitial detection27,127,458Covering all potential variation sites in the genomeBasic quality control8,946,312Eliminate low-quality and multi-allelic loci to ensure the accuracy of typingPopulation-level screening2,798,453Exclude low-frequency and high-missing-site variants to enhance the statistical test efficacySite quality screening2,789,615Further filter out the low-confidence lociLD trimming116,024Reduce locus linkage redundancy and ensure the independence of population structure analysis Table 2Summary of SNP filtering and quality control steps.Screening stepsThe remaining number of SNPsSelection criteriaInitial detection27,127,458Covering all potential variation sites in the genomeBasic quality control8,946,312Eliminate low-quality and multi-allelic loci to ensure the accuracy of typingPopulation-level screening2,798,453Exclude low-frequency and high-missing-site variants to enhance the statistical test efficacySite quality screening2,789,615Further filter out the low-confidence lociLD trimming116,024Reduce locus linkage redundancy and ensure the independence of population structure analysis Summary of SNP filtering and quality control steps. SNP annotation analysis and variant detection SNP annotation in this study was based on a dataset of 2,789,615 high-quality SNPs (loci without LD pruning after quality control). The average sequencing depth per sample was 12.37×, with an average alignment rate of 99.65% and an average Q30 of 95.07%. The GC content remained stable within the range of 44.0% to 44.6%. All quality control metrics met the standards for whole-genome resequencing studies. The annotation results showed that among the 27,127,458 original SNPs, SNPs located in the exon regions accounted for 3.68% (including 94,045 synonymous variants and 66,960 non-synonymous variants). SNPs in the intron regions made up 28.89%, and intergenic SNPs accounted for 68.94%. The transition/transversion ratio (ts/tv) was 2.484, which is consistent with the typical characteristics of mammalian genomic variations. The chromosomal distribution of SNP loci is shown in Fig. 1.Fig. 1Heatmap of SNP Distribution. The density of SNPs within each 0.1 Mb window was calculated. The vertical axis represents chromosomes, while the horizontal axis indicates positions along each chromosome. Redder colors at specific positions indicate higher numbers of variant sites. Heatmap of SNP Distribution. The density of SNPs within each 0.1 Mb window was calculated. The vertical axis represents chromosomes, while the horizontal axis indicates positions along each chromosome. Redder colors at specific positions indicate higher numbers of variant sites. Population structure analysis Based on the 116,024 independent SNP markers obtained after quality control filtering, population structure analysis was performed. The phylogenetic tree showed a clear clustering pattern that divided the population into four subgroups (Fig. 2a). Principal component analysis (PCA) revealed that the first principal component (PC1) accounted for 18.2% of the total genetic variation among individuals, while PC2 explained 11.5%, together representing 29.7% of the overall genetic diversity in the studied population (Fig. 2b). The PCA plot indicated that the majority of samples clustered into four distinct regions, with only a few individuals deviating from the main groups. The optimal number of clusters was determined to be 4 (Fig. 2c). When stratified according to K = 4, the genetic components of the four subgroups could be clearly distinguished (Fig. 2d). These results showed good correspondence among the phylogenetic tree, PCA, and Admixture analyses.Fig. 2Population structure analysis of male Dongfeng sika deer. (a) Systematic evolution tree; (b) PCA analysis; (c) Optimal K-group analysis; (d) Population structure diagram. Population structure analysis of male Dongfeng sika deer. (a) Systematic evolution tree; (b) PCA analysis; (c) Optimal K-group analysis; (d) Population structure diagram. GWAS of sika deer body weight traits based on the MLM model The results of the GWAS analysis for the body weight trait of sika deer using the MLM model yielded the Manhattan plot and quantile-quantile (Q-Q) plot for the body weight of male sika deer, as shown in Fig. 3. The Q-Q plot analysis revealed that after correcting for population stratification using the first three principal components (PC1–PC3), the observed P-values overlapped with the expected P-values in most intervals (λ = 1.03), with deviations only at the far-right end of the distribution. This indicated that the confounding effects of population stratification had been effectively controlled, and the association signals primarily originated from true SNP-body weight associations rather than interference from confounding factors. The details of the extremely significant SNP loci and their related genes are presented in Table 2. The SNP locus chr24:55874319 on chromosome 24 showed the strongest association with body weight traits, with the highest phenotypic variance explained (PVE) of 2.3%. This locus is located in the exon region of the ERC2 gene and serves as the core candidate locus influencing the body weight trait of male Dongfeng sika deer. The PVE of the remaining significant SNPs ranged from 0.4% to 1.8%, all contributing to the phenotypic variation to a certain extent. In addition to chromosome 24, the MLM model also detected loci on chromosomes 9, 12, 13, and 14, which were annotated to the LOC122708028, FHIT, CASP14, UBE3A, NRXN3, and GABRA5 genes. This suggests that these gene regions may contain important information or regulatory elements related to body weight development Table 3.Fig. 3Genome-wide association study results for body weight in male Dongfeng sika deer. (a) Manhattan plot; (b) Q-Q plot.Table 3Highly significant SNP loci and related genes from GWAS based on the MLM model.ChromosomePosition, bpPNearby genesPosition2455,874,3191.81E-19ERC2Intragenic142,063,5062.94E-19LOC122708028Upstream2452,547,8339.93E-17FHITIntragenic911,967,9221.43E-15CASP14Downstream911,967,9121.43E-15CASP14Downstream136,800,8731.82E-15UBE3ADownstream911,967,8999.91E-15CASP14Downstream911,967,9231.37E-14CASP14Downstream1212,441,7471.07E-13NRXN3Intragenic911,967,9322.04E-13CASP14Downstream911,963,8145.09E-13CASP14Downstream138,442,7302.04E-12GABRA5Intragenic Genome-wide association study results for body weight in male Dongfeng sika deer. (a) Manhattan plot; (b) Q-Q plot. Table 3Highly significant SNP loci and related genes from GWAS based on the MLM model.ChromosomePosition, bpPNearby genesPosition2455,874,3191.81E-19ERC2Intragenic142,063,5062.94E-19LOC122708028Upstream2452,547,8339.93E-17FHITIntragenic911,967,9221.43E-15CASP14Downstream911,967,9121.43E-15CASP14Downstream136,800,8731.82E-15UBE3ADownstream911,967,8999.91E-15CASP14Downstream911,967,9231.37E-14CASP14Downstream1212,441,7471.07E-13NRXN3Intragenic911,967,9322.04E-13CASP14Downstream911,963,8145.09E-13CASP14Downstream138,442,7302.04E-12GABRA5Intragenic Highly significant SNP loci and related genes from GWAS based on the MLM model. Gene function enrichment analysis results The results of KEGG enrichment analysis for body weight trait in sika deer are presented in Table 4, with the pathway visualization shown in Fig. 4. Among the enriched genes in KEGG pathways, the GABRA5 and GABRG3 were significantly associated with the γ-aminobutyric acid neurotransmission pathway; the S1PR1 gene was involved in the 1-phosphatidylinositol signaling pathway; and the TSHR gene was involved in the thyroid-stimulating hormone signaling pathway20.Fig. 4KEGG enrichment analysis plot.Table 4KEGG enriched pathways related to body weight trait in sika deer.IDMolecular functionP-valueNumber of genesGenehsa04080Neuroactive ligand-receptor interaction0.0134GABRA5, S1PR1, GABRG3, TSHRhsa05033Nicotine addiction0.0562GABRA5, GABRG3 KEGG enrichment analysis plot. Table 4KEGG enriched pathways related to body weight trait in sika deer.IDMolecular functionP-valueNumber of genesGenehsa04080Neuroactive ligand-receptor interaction0.0134GABRA5, S1PR1, GABRG3, TSHRhsa05033Nicotine addiction0.0562GABRA5, GABRG3 KEGG enriched pathways related to body weight trait in sika deer. The results of GO enrichment analysis for body weight trait in sika deer are presented in Table 5, with the pathway visualization illustrated in Fig. 5. Multiple genes were found to participate in signaling and regulatory pathways. Specifically, GABRA5 and GABRG3 were involved in the GABAergic neurotransmission pathway; OLFM3 participated in the Wnt/β-catenin signaling pathway; and RPS6KA5 was associated with the MAPK/ERK/p38 signaling pathway. The FOXN3 gene was implicated in the cell-cycle checkpoint pathway, while SAP30 was implicated in the Sin3 transcriptional co-repression pathway. Additionally, RAP2B was associated with the MAPK pathway, and TMEM63C was associated with the mechanosignaling pathway. The VEPH1 gene participated in the Hippo-YAP/TAZ pathway; S1PR1 was involved in the 1-phosphatidylinositol signaling pathway; TTC7B participated in the cholesterol absorption and homeostasis regulation pathway; and FHIT was associated with the DNA damage repair and cell cycle regulation pathway.Fig. 5GO enrichment analysis plot. This figure is adapted from the KEGG pathway database (Kanehisa Laboratories, https://www.kegg.jp).Table 5GO enriched pathways related to body weight trait in sika deer.IDMolecular functionP-valueNumber of genesGeneCategoryGO:0060078Regulation of postsynaptic membrane potential0.0122GABRA5, GABRG3BPGO:0051932Synaptic transmission, GABAergic0.0122GABRA5, GABRG3BPGO:0007214Gamma-aminobutyric acid signaling pathway0.0202GABRA5, GABRG3BPGO:0006821Chloride transport0.0312GABRA5, GABRG3BPGO:0034220Monoatomic ion transmembrane transport0.0562GABRA5, GABRG3BPGO:0050877Nervous system process0.0562GABRA5, GABRG3BPGO:0007165Signal transduction0.0823OLFM3, GABRA5, GABRG3BPGO:0006355Regulation of DNA-templated transcription0.0853RPS6KA5, FOXN3, SAP30BPGO:0032590Dendrite membrane0.0172GABRA5, GABRG3CCGO:1,902,711GABA-A receptor complex0.0202GABRA5, GABRG3CCGO:0098794Postsynapse0.0262GABRA5, GABRG3CCGO:0005886Plasma membrane0.0326RAP2B, TMEM63C, VEPH1, S1PR1, TTC7B, FHITCCGO:0008503Benzodiazepine receptor activity0.0142GABRA5, GABRG3MFGO:0022851GABA-gated chloride ion channel activity0.0142GABRA5, GABRG3MFGO:0005237Inhibitory extracellular ligand-gated monoatomic ion channel activity0.0142GABRA5, GABRG3MFGO:0004890GABA-A receptor activity0.0262GABRA5, GABRG3MF GO enrichment analysis plot. This figure is adapted from the KEGG pathway database (Kanehisa Laboratories, https://www.kegg.jp). Table 5GO enriched pathways related to body weight trait in sika deer.IDMolecular functionP-valueNumber of genesGeneCategoryGO:0060078Regulation of postsynaptic membrane potential0.0122GABRA5, GABRG3BPGO:0051932Synaptic transmission, GABAergic0.0122GABRA5, GABRG3BPGO:0007214Gamma-aminobutyric acid signaling pathway0.0202GABRA5, GABRG3BPGO:0006821Chloride transport0.0312GABRA5, GABRG3BPGO:0034220Monoatomic ion transmembrane transport0.0562GABRA5, GABRG3BPGO:0050877Nervous system process0.0562GABRA5, GABRG3BPGO:0007165Signal transduction0.0823OLFM3, GABRA5, GABRG3BPGO:0006355Regulation of DNA-templated transcription0.0853RPS6KA5, FOXN3, SAP30BPGO:0032590Dendrite membrane0.0172GABRA5, GABRG3CCGO:1,902,711GABA-A receptor complex0.0202GABRA5, GABRG3CCGO:0098794Postsynapse0.0262GABRA5, GABRG3CCGO:0005886Plasma membrane0.0326RAP2B, TMEM63C, VEPH1, S1PR1, TTC7B, FHITCCGO:0008503Benzodiazepine receptor activity0.0142GABRA5, GABRG3MFGO:0022851GABA-gated chloride ion channel activity0.0142GABRA5, GABRG3MFGO:0005237Inhibitory extracellular ligand-gated monoatomic ion channel activity0.0142GABRA5, GABRG3MFGO:0004890GABA-A receptor activity0.0262GABRA5, GABRG3MF GO enriched pathways related to body weight trait in sika deer. Discussion The core experimental population of this study consisted of 266 male Dongfeng sika deer. Although the sample size was smaller than the conventional scale for GWAS in livestock and poultry, it was constrained by species‑specific characteristics and practical conditions: sika deer exhibit strong wildness, requiring professional anesthesia for sampling, and deer farms are concerned that sampling may affect antler production. Under these circumstances, 266 individuals already represent a relatively extensive coverage of the core breeding population, capable of reflecting its core genetic background. In this study, high‑quality data were obtained via whole‑genome resequencing, with over 2.79 million high‑quality SNPs screened through multi‑dimensional quality control. Combined with the MLM model incorporating multiple corrections for population stratification and genetic relatedness (λ = 1.03), the limitation of small sample size was effectively compensated for. Body weight traits in sika deer serve as crucial indicators for evaluating their production performance and economic value. As highly significant phenotypic characteristics in biological genetics and breeding processes, they play a pivotal role in selection, breeding, and husbandry management21. The Dongfeng sika deer, in particular, demonstrates notable advantages, including strong disease resistance, high-yielding, high-quality antlers, and pronounced hybrid vigor, making it an excellent genetic resource for meat deer breeding programs22,23. In the present study, a GWAS was conducted on 266 Dongfeng sika deer bucks to identify SNP loci associated with body weight variation. The objective was to screen potential molecular markers for marker-assisted selection and to explore functional genes that may play regulatory roles in body weight determination. These findings provide a robust theoretical foundation and technical support for optimizing genetic improvement strategies for body weight in Dongfeng sika deer stags, thereby enhancing the efficiency of genetic selection in sika deer breeding. Population structure is a crucial characterization of a species’ genetic diversity. Its analytical results not only provide a key basis for correcting confounding effects in genome-wide association studies (GWAS) but also reflect the breeding history and genetic differentiation characteristics of a breed5,23. Based on 116,024 independent SNP markers, this study comprehensively confirmed through phylogenetic tree construction, principal component analysis (PCA), and Admixture analysis that the optimal number of clusters for the population of 266 male Dongfeng sika deer was K = 4. The four subgroups exhibited clear genetic boundaries, which were highly consistent with the clustering results of the phylogenetic tree (Fig. 2), indicating that distinct genetic stratification existed within this population. This result is closely related to the breeding background of the Dongfeng sika deer. As the “Hometown of Sika Deer in China”, Dongfeng County has long maintained purebred breeding of multiple strains and isolated feeding of local subpopulations. Different subgroups may correspond to different core breeding families or feeding units. Long-term directional breeding coupled with limited gene flow has jointly shaped the current population genetic structure1. From the perspective of GWAS reliability, population stratification is one of the primary factors leading to false positives in association analysis19. The genetic background differences among the four subgroups in this study (Fig. 2d) might cause inter-group variation in body weight phenotypes. Without proper correction, such subgroup differences could easily be misjudged as the genetic effects of SNP loci. Therefore, in this study, the first three principal components (PC1–PC3) from PCA were included as fixed effects in the mixed linear model (MLM), combined with the kinship matrix (K matrix) to correct for genetic relatedness among individuals. Ultimately, verification via the Q-Q plot (λ = 1.03) ensured the authenticity of the association signals. This approach is consistent with the mainstream strategies for population structure correction in GWAS studies of livestock and poultry, such as cattle and sheep2. Using the MLM approach, the GWAS identified twelve loci with highly significant associations with body weight traits in Dongfeng sika deer stags. These significant SNPs were annotated to seven candidate genes: ERC2, LOC122708028, FHIT, CASP14, UBE3A, NRXN3, and GABRA5. Notably, these seven genes do not function in isolation but rather constitute an interactive regulatory network. Specifically, ERC2 primarily participates in energy metabolism signaling and the regulation of cell proliferation and growth, suggesting its possible role in tissue development and metabolic balance. LOC12270802824 influences adipocyte differentiation and muscle growth-related genes, contributing to variations in body mass composition. FHIT25 is involved in maintaining the balance in cell proliferation and in modulating indirect energy metabolism. CASP1426 plays a role in the regulation of adipose tissue inflammation and the association between adipose tissue and energy reserves. UBE3A27 is implicated in the degradation of metabolism-related proteins and the regulation of the neuro-metabolic axis. Furthermore, NRXN3 participates in neural regulation of appetite and energy balance, as well as in metabolic hormone signaling. GABRA528 functions within the central nervous system’s inhibitory pathways, regulating appetite suppression and energy expenditure. The synergistic interplay among these genes bridging neural circuits (GABRA5, NRXN3), metabolic regulation (ERC2, LOC122708028, CASP14), and cellular homeostasis (FHIT, UBE3A) provides potential clues for deciphering the genetic basis of body weight traits in sika deer through multi-network regulation involving the “neuro-metabolic-cellular system.” To further explore the biological functions and potential regulatory networks of the genes associated with body weight, we performed GO and KEGG enrichment analyses on the genomic regions surrounding the significant SNPs (including the seven GWAS-identified genes and their linked loci). This approach identified significantly enriched pathways and functional terms, which extended to additional interacting genes that were not directly identified as GWAS candidates but co-occurred in the same regulatory pathways. These results systematically reveal the potential cooperative mechanisms of 13 genes and enhance understanding of the genetic regulatory network underlying body weight in sika deer, establishing a foundation for subsequent functional verification and genetic improvement programs. Among the identified genes, GABRA5 and GABRG328 form the “neuro-regulatory core.” Both genes were significantly enriched in GABA-gated chloride ion channel activity and postsynaptic membrane potential regulation, and were involved in the GABAergic neurotransmission pathway (all P < 0.05). During embryogenesis, these genes promote neural progenitor cell proliferation and brain region formation, while postnatally they regulate appetite and energy metabolism via the hypothalamus and participate in the neuroactive ligand-receptor interaction pathway (P < 0.05) together with S1PR129 and TSHR30. Although their involvement in the nicotine addiction pathway (P > 0.05) showed only marginal significance, it suggested a limited association with growth and development. TSHR30, as a “key endocrine regulator,” participates in the thyroid-stimulating hormone signaling pathway (P < 0.05). By maintaining T3 and T4 stability via the hypothalamic-pituitary-thyroid axis, TSH regulates systemic metabolism, skeletal growth (stimulating chondrocyte proliferation and epiphyseal development), and neuronal myelination. Importantly, TSHR’s interaction with the GABAergic pathway highlights crosstalk between neural and endocrine systems. Additionally, OLFM331 aids signal transduction and enhances neural network formation by promoting synaptogenesis. Although its involvement in the Wnt/β-catenin signaling pathway is marginally significant (P > 0.05), RPS6KA5, FOXN3, and SAP30 collectively maintain “transcriptional homeostasis maintenance.” Specifically, RPS6KA532 activates growth-related gene expression through histone modification and transcriptional regulation with a marginal association with the MAPK/ERK/p38 signaling pathway (P > 0.05). FOXN333 governs cell proliferation and preserves genomic stability through pathways such as cell cycle checkpoints and p53-mediated DNA damage responses. SAP3034 regulates transcriptional repression via histone deacetylation within the Sin3 co-repressor complex. Together, these genes ensure precise transcriptional regulation, which is essential for normal cellular function and organismal growth. At the cellular level, RAP2B, TMEM63C, VEPH1, S1PR1, TTC7B, and FHIT were enriched at the cell membrane (P < 0.05), collectively forming the “cellular foundation” for growth and development. RAP2B35 activates the MAPK/ERK pathway to promote embryonic organ primordium formation, while TMEM63C36 regulates orderly cell cycle division via calcium signaling. VEPH137 maintains cytoskeletal stability to facilitate neuronal migration and tissue morphogenesis. S1PR129 regulates angiogenesis and hematopoietic stem cell homing through the PI3K/Akt signaling pathway and is significantly involved in the 1-phosphatidylsphingosine signaling pathway (P < 0.05). TTC7B38 plays a role in growth factor processing and secretion, ensuring effective signal transduction; it also contributes to cholesterol absorption and homeostatic regulation. Meanwhile, FHIT25 prevents abnormal cell proliferation by modulating DNA damage repair and activating apoptosis pathways. Through a synergistic network encompassing “cellular foundation support-neuro-endocrine core regulation-transcriptional homeostasis maintenance,” these genes span critical stages, including embryonic organogenesis, neural maturation, and somatic growth. This provides comprehensive targets and theoretical foundations for elucidating the molecular mechanisms underlying the growth and development of body weight traits in sika deer, while also addressing gaps in current research on genes and pathways significantly associated with body weight traits in sika deer. Conclusions The results of this study are consistent with the original research objectives, which aimed to identify key molecular markers and candidate genes for body weight traits in Dongfeng sika deer using genome-wide association analysis. We detected 12 significant SNPs annotated to seven important genes: ERC2, LOC122708028, FHIT, CASP14, UBE3A, NRXN3, and GABRA5. The most significant locus was located in the ERC2 gene. Functional enrichment indicated that these genes are mainly involved in GABAergic synaptic transmission, DNA damage repair, and cell cycle regulation pathways. These findings provide effective molecular markers for body weight evaluation and molecular breeding in sika deer. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1