Background & Summary

The domestic goat is an important economic livestock in developing countries, especially in Asia and Africa, as it provides essential products such as milk, meat, and mohair, cashmere, and furs produced from cashmere goats1. China ranks first in the world both in the number of breeds and the total population of goats raised2. In 2024, 69 national goat breeds were recognized by the National Catalogue for Livestock and Poultry Genetic Resources (http://www.zys.moa.gov.cn/gsgg/), with numbers concentrated in mainly South China. Beyond these national-level goat breeds, numerous local breeds exist in smaller populations. Guizhou province, featuring a subtropical mountain climate, is the center of the Karst ecosystems in Southwest China3. The province is home to several goat breeds, including but not limited to the Guizhou black goat (GBG), the Hezhang black goat (HBG), and the Tashi goat (TG). Through long-term natural and artificial selection, these breeds have developed small body size and strong mountain foraging ability, well-suited to the local Karst ecosystem and providing a basic income for farmers.

As the GBG is the leading goat breed in Guizhou Province, it has gained extensive research attention, particularly regarding its growth performance. For example, Yuan et al.4 investigated the effect of allicin on the growth performance of GBG, while Long et al.5 focused on the effect of Chinese herbal medicine residues. Based on variant-trait associations, the Insertions/Deletions (InDels) in GATA46, and the copy number variants in CADM27, Opn48, SNX299, and MYLK410 were identified to be significantly associated with growth traits; the Single Nucleotide Polymorphisms (SNP) in ACADM11 was determined to be significantly associated with slaughter and meat quality traits. Additionally, this breed’s reproductive performance and population structure have generated some attention. Before this study, the largest genome resequencing dataset of GBG, comprising 30 individuals, was conducted by Prof. Wang’s team12,13 and by integrating this dataset and other downloaded data, they performed a population structure and selective signals analyses to explore genomic characteristics12 and causal structural variants linked to mutton flavor13. Compared to GBG, research on HBG and TG remains scarce, and for HBG, no relevant publications were obtained from the Web of Science or the PubMed database. For TG, there is only one study from this lab that identified genomic variants associated with body conformation traits14. Overall, the absence of sufficient genomic information for these local breeds severely hampers breed genetic evaluation, functional gene dissection, genetic improvement, and conservation efforts. Hence, additional genome sequence data is desperately needed.

Whole-genome sequencing (WGS) is a powerful technique that enables researchers to analyze an organism’s entire genetic makeup, and a considerable number of studies used WGS data for different reasons. For example, Cai et al.15 investigated the evolutionary history of cashmere-producing goats in China by integrating ancient and current goat genome sequencing data, while Liu et al.16 constructed a goat pan-genome using the WGS dataset that comprises 813 individuals and revealed the patterns of gene loss during domestication. Based on a genome-wide association study, WGS was used to identify genes associated with traits such as milk yield17, cashmere yield18, hair diameter19, and hair density20. Based on a selective sweep analysis, some breed-specific traits have been elucidated, such as the high-altitude adaptation of Tibetan cashmere goats21 and the rapid muscle growth of Boer goat22. Additionally, the WGS technique has also been used to evaluate breed genetic characteristics23,24 and provide core variants to develop SNP chips25,26. Altogether, WGS offers comprehensive genetic insights essential for evolutionary study, trait analysis, breeding, and conservation. Consequently, this makes the generation of genomic data for underrepresented breeds increasingly essential.

Here, we present a new WGS dataset from three indigenous goat breeds, including GBG (n = 104), HBG (n = 100), and TG (n = 100). The dataset, encompassing 6.0 TB of raw sequence data, constitutes the largest WGS dataset generated from the Karst region of Southwest China to date. Sequencing was performed at an average depth of 7.5X, ensuring the necessary power and resolution for genomic analyses. By aligning the sequencing data to the Capra hircus reference genome (ARS1.227), https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_001704415.2/) and performing variant calling and variant filtration, a total of 27.13 million SNPs and 2.76 million InDels were identified. The reliability of this WGS dataset is evident from its sequencing base quality, variant quality, sample independence, and breed independence.

This dataset will fill gaps in the genomic resources of these goat breeds, allowing for the (1) Calculation of genetic metrics to evaluate their current status, and infer their genetic relationships; (2) Identification of genomic variants associated with biological traits; (3) Integration of other genomic resources and tracing species evolution and domestication; (4) Comparison of genomic data of different breeds to identify regions under positive selection; and (5) Use of core variants to develop SNP chips for future breeding purposes. Altogether, this large-scale WGS dataset from the Karst region of Southwest China significantly enriches global goat breed genome resources and is crucial for studying population genetics and elucidating economic traits.

Methods

Animals and sample collection

This experiment was approved by the Animal Care and Use Committee of Guizhou University (No.EAE-GZU-2023-E047). A total of 304 adult goats, including Guizhou black goats (GBG, male n = 84, female n = 20), Hezhang black goats (HBG, male n = 80, female n = 20), and Tashi goats (TG, male n = 19, female n = 81), were collected from Guizhou province, China (Fig. 1a). To thoroughly sample the genetic diversity of goat breeds, individuals were selected from dozens of smallholder farms in Anshun, Bijie, and Rongjiang City, respectively (Fig. 1b). These breeds exhibit distinct phenotypes, such as differences in coat color and horn type (Fig. 1c). The goats were allowed to graze naturally and were supplemented with corn and dry hay. Then, 3–5 mL jugular venous blood was sampled from each animal, anticoagulated with EDTA, and stored at −20 °C until DNA extraction.

Fig. 1
Fig. 1
Full size image

Sampling locations and study workflow. (a) The three goat breeds included the Guizhou black goat (GBG, n = 104), the Hezhang black goat (HBG, n = 100), and the Tashi goat (TG, n = 100). (b) Sampling area and geographic distribution of the study populations. (c) Breed characteristics in terms of phenotype and production, and the purpose of the breed. (d) Workflow of this study.

Whole genome sequencing

The DNA was extracted from blood samples of 304 goats using the standard phenol-chloroform protocol (Fig. 1d). The concentration, integrity, and purity of the genomic DNA were assessed using agarose gel electrophoresis and a NanoDrop spectrophotometer (Thermo Scientific, Waltham, MA, USA). During library construction, the genomic DNA was amplified and randomly broken into fragments of about 150 bp, followed by the addition of sequencing adapters. Then, the DNA library was sequenced by Compson Biotechnology Company (Beijing, China) using the BGI-T7 platform.

Genomic alignment and variant calling

The genomic analysis pipeline primarily comprised quality control, read mapping, variant calling, variant filtering, and variant annotation (Fig. 1d). The WGS raw data in fastq files for all 304 goats, were quality-controlled using the fastp software (v0.23.428) to obtain clean data, while sequence alignment and variant detection were processed using the Sentieon Genomics software (v20230829). Briefly, clean reads were aligned to the goat reference genome (ARS1.227), whereafter the BAM files were sorted and duplicates marked. Variant calling was conducted using the Sentieon haplotyper module to independently generate a genomic Variant Call Format (gVCF) file for each individual. Finally, variant joint calling was performed in the Sentieon GVCFtyper module to create a common VCF file from all the gVCF files. It is important to note that the module tools in Sentieon software are fully faithful to the classic BWA-GATK pipeline30,31.

Variant types and variant annotation

Statistical metrics generated in the above variant calling step, include Mapping Quality (MQ), Quality by Depth (QD), Fisher Strand (FS), and Strand Odds Ratio (SOR), evaluating coverage depth and alignment quality at variant positions, which help filter out potential false positives and ensure accurate variant calling. Both SNP and InDel variants were filtered using the SelectVariants module in the GATK software (v4.1.8.131). Further filtration was performed using the Vcftools software (v0.1.1632) to filter out variants when the average depth was less than 5 and the missing genotype rate exceeded 10% in all samples. The variant depth and cumulative proportions were performed to assess variant depth, which in turn serves as an indicator of variant quality, while the filtered SNPs and InDels were functionally annotated using the snpEff software (v.5.133). Additionally, the variant locations in intronic, untranslated, upstream, downstream, and intergenic regions were calculated.

Principal component analysis, genetic kinship, and population structure

To evaluate sample and breed independence, the principal component analysis (PCA), kinship analysis, phylogeny analysis, and population structure analysis were conducted based on genome-wide SNPs. The PCA was performed in Plink (v0.7634), and each principal component was tested based on the twstat method using the EIGENSTRAT software (v6.1.435). The kinship matrix was calculated in the GEMMA software (v0.98.536) and visualized in the R package heatmap (v1.0.12). Phylogenetic distance was estimated using the VCF2Dis software (v1.54, https://github.com/BGI-shenzhen/VCF2Dis) and visualized using Figtree software (v1.4.4, http://tree.bio.ed.ac.uk/software/figtree/). Under the different hypothetical subpopulations (from K = 2 to K = 10), the population structure for these samples was calculated using the Admixture software (v1.3.037) and visualized in the Python package PONG (v1.538).

Data Records

All original genome sequencing data in FASTQ format have been deposited in the Genome Sequence Archive39 on the China National Center for Bioinformation (CNCB) platform under accession number CRA025744 (https://ngdc.cncb.ac.cn/gsa/browse/CRA025744)40, and Sequence Read Archive on National Center for Biotechnology Information (NCBI) under accession number PRJNA1281799 (https://www.ncbi.nlm.nih.gov/sra/SRP594418)41. The final SNP.vcf (27,134,128 SNPs) and InDel.vcf (2,762,642 InDels) files were deposited in the Genome Variation Map42 on the CNCB platform under accession number GVM00105143, and the European Variation Archive (EVA)44 under accession number PRJEB9083145.

Technical Validation

Quality control for sequencing data

The whole genome sequencing of 304 goats using the BGI-T7 platform yielded 5968.2 Gb of raw sequence data (Fig. 2a, Table 1). For each sample, the unfiltered reads ranged from 97 to 386 million reads, and sequencing yields ranging between 11.3 Gb and 57.9 Gb were obtained. The average sequencing depth was 7.5X for all the samples, which varied from 5.0X to 22.0X (Fig. 2a, Table 1). For all samples, 97.4–99.4% and 93.2–98.0% of the bases achieved the high Phred quality score of Q20 (sequencing error rate < 0.01) and Q30 (sequencing error rate < 0.001), indicating the high base calling accuracy, respectively. Besides, the average GC content varied from 39.7% to 44.4%. Each position in the 150 base pair read obtained high-quality scores of 35 (Fig. 2b), and almost all the reads had quality scores that varied from 35 to 40, confirming the overall high-quality scores of all the sequencing reads (Fig. 2c). For all samples, the genome coverage of the sequencing reads on the goat reference genome exceeded 95% (Fig. 2d), the mapped depth exceeded 5X (Fig. 2e), and the properly mapped rates varied from 73.54% to 99.2% (average of 97.0%). These indicators demonstrate the high quality of the sequencing data based on sequencing data yield, base quality, and genome coverage.

Fig. 2
Fig. 2
Full size image

Statistics of sequencing data and read alignment. (a) The sequencing yield, reads, depth, phred quality scores (Q20 and Q30), GC content, and mapping rate of high-throughput sequencing data of 304 goats. (b) Average quality score per base in a 150-base-pair read. (c) The sequence quality score plot revealed that nearly all reads have quality scores ranging from 35 to 40. (d) The genome coverage of sequencing reads on the 29 autosomes of the goat reference genome exceeded 95%. (e) The mapped depth of sequencing reads on the 29 autosomes of the goat reference genome. The mapped depth of all samples exceeded 5X. Each colorful circle or line represents one sample of the 304 goats.

Table 1 Whole Genome Sequencing data statistics of 304 goats from the Karst region in China.
Full size table

Quality control of SNP and InDel data

After performing joint calling for all samples, a total of 41,559,429 SNPs and 4,891,077 InDels were obtained. To ensure variant quality and minimize false positives, we performed hard filtration using the GATK software31 and assessed variant quality using statistical metrics including MQ, QD, FS, and SOR, obtaining 31,405,939 SNPs and 4,603,254 InDels. The Vcftools software (v0.1.1632) filtered out variants whose average depth was less than 5 and whose missing genotype rate exceeded 10%. Finally, a total of 27,134,128 SNPs and 2,762,642 InDels were retained (Fig. 3a,b).

Fig. 3
Fig. 3
Full size image

Variant locations and their cumulative distribution using WGS data of 304 goats. Proportions of all filtered SNPs (a) and InDels (b) at specific chromosomal location categories. The cumulative depth distribution of SNPs (c) and InDels (d) in each of the 304 samples. The average depths for SNPs and InDels were 7.2 and 7.5, respectively. Each colorful line represents one of the 304 goats.

Summary statistics of SNPs and InDels

High-quality variants were evenly dispersed throughout the 29 autosomes of the goat genome, with an average frequency of one SNP per 85 bases and one InDel per 838 bases (Table 2). As shown in Fig. 3a,b, more than half of the SNPs (54.3%) and InDels (56.1%) were located in intronic regions, while only a small percentage (0.8% of SNPs and 0.2% of InDels) were located in Exons. Approximately 1% of variants were located in UTR regions, although the rate of total numbers of SNPs and InDels was approximately 10:1, their distribution and variant classes were similar (Fig. 3a,b). The cumulative depth distribution plots (Fig. 3c,d) illustrated the cumulative proportional curves of variant depth from 1 to 40, and the arithmetic average of variant depth for all SNPs and InDels was 7.2 and 7.5, respectively. These results point to the uniform distribution and high quality of the filtered variants.

Table 2 Statistics of the final SNPs and InDels identified from each chromosome in the WGS data of three goat breeds in China.
Full size table

Sample and breed independence

We conducted population structure analysis using PCA, kinship analysis, and phylogenetic tree analysis to assess both sample-level and breed-level genetics independence. The first two principal components explained 15.3% of the total variation and showed a clear distinction between the three goat breeds (Fig. 4a). Additionally, the kinship matrix heatmap showed almost no kinship between breeds, with low kinship coefficients within each breed (Fig. 4b). The histogram frequency distribution showed that 90% and 95% of the relatedness coefficients were lower than 0.037 and 0.093, respectively, indicating low kinship among samples and suggesting good sample independence (Fig. 4c). The phylogenetic tree (Fig. 4d) showed the evolutionary relationship of the three goat breeds, and while the three breeds were relatively independent, TG and HBG showed some influence from GBG. As shown in Fig. 4e, the lowest cross-validation error value was 0.268 at K = 3. Population structure analysis revealed that the ancestral composition of all samples can be easily distinguished by breed at K = 3 (Fig. 4f), and although a few samples showed ancestry from other breeds, this is considered normal given their geographical proximity of 200 km. These results effectively demonstrated sample independence, breed independence, and the high quality of the whole-genome resequencing dataset.

Fig. 4
Fig. 4
Full size image

Population genetic analyses using WGS data from 304 goats indicated satisfactory sample independence and breed independence. (a) PCA reveals population stratification among the three goat breeds. (b) The kinship matrix calculated based on genome-wide SNPs can also differentiate the three goat breeds. (c) The histogram frequency distribution of the pairwise kinship coefficients, where low kinship coefficients among most samples indicate good sample independence. (d) Maximum-likelihood phylogeny was constructed using genome-wide SNPs of 304 goats. (e) The line chart of cross-validation (CV) error under different numbers of hypothetical subpopulations. (f) ADMIXTURE analysis of 304 samples based on genome-wide SNPs under a model with three ancestral components (K = 3). These results show breed independence and low intra-breed kinship, confirming the high quality of the WGS dataset. GBG, Guizhou black goat; HBG, Hezhang black goat; TG, Tashi goat.

Usage Notes

This study presents both original raw reads and processed variant files. Notably, these variants were obtained based on the most common reference genome of the San Clemente breed (ARS1.2, https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_001704415.2/). Our dataset includes both bucks and does, but due to the lack of effective Y-chromosome information in the reference genome, variants on this chromosome were missing in the VCF files. The other two telomere-to-telomere genome assemblies of the goat genome are available in the NCBI database: ASM4082201v1 (https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_040822015.1/), assembled by Prof. Su’s team46, and T2T-goat2.0 (https://www.ncbi.nlm.nih.gov/datasets/genome/GCA041735815.1/), assembled by Prof. Li’s team47. These genome versions offer greater flexibility for utilizing our WGS dataset in future applications.