A genomic approach for accurate identification of closely related species with next-generation sequencing samples

Abstract

Accurate species identification from Whole Genome Sequencing (WGS) data remains challenging, particularly for closely related species such as sheep (Ovis aries) and goats (Capra hircus). Through analysis of mapping quality metrics and Kraken2 taxonomic classification of 40 WGS sheep and goat samples, we demonstrate that conventional approaches yield ambiguous results, with overlapping alignment rates and inconclusive taxonomic assignments. We present a robust comparative genomic approach that uses species-specific genomic regions to distinguish these species in WGS samples. We define species-specific regions as those exhibiting distinctive coverage patterns: average coverage when samples are aligned to their matching reference genome but absent/low coverage when aligned to non-matching references. By analyzing WGS data from both species aligned to both reference genomes, we identified 155,800 goat-specific and 1,714,126 sheep-specific regions. After curation, 10 high-confidence regions per species were selected, achieving 100% accuracy within the analyzed validation datasets comprising 14 independent samples. This approach provides reliable species verification for WGS data and establishes a framework that could be extended to other closely related species. To facilitate adoption, we provide analysis scripts and curated genomic regions available on our GitHub repository (https://github.com/BTC-Lab/Goat_Sheep_specific_regions).

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-41497-0.

Introduction

Accurate species identification in unlabeled next-generation sequencing (NGS) samples is a cornerstone of many biological and agricultural studies, ensuring that downstream analyses are both valid and reliable^1–3. However, identifying species becomes particularly challenging when dealing with closely related species due to their high genomic similarity^2,4. In such cases, sequencing reads from one species might align just as well to the reference genome of another, leading to ambiguous or even misleading results. These challenges can complicate critical research areas, such as population genetics, genomic studies, and trait mapping, where precise species assignment is essential².

Advances in genomics and bioinformatics have accelerated biological sciences, making Whole Genome Sequencing (WGS) more accessible. However, many public NGS datasets lack complete or well-documented metadata, making it difficult to determine sample origins^5,6. Human errors in sample collection, labeling, and processing can lead to mislabeled datasets, resulting in flawed conclusions that impact agricultural and conservation practices.

Conventional sequence alignment tools like BLAST⁷ and taxonomic classification programs like Kraken2⁸ rely on sequence homology to reference databases. While effective in many scenarios, these methods often fall short when tasked with distinguishing closely related species^9,10. This limitation underscores the need for new strategies that go beyond simple alignment to achieve more reliable species identification.

Single nucleotide polymorphisms (SNPs) are commonly used as genetic markers but pose challenges in distinguishing closely related species due to issues with variant calling, particularly for rare variants or low-coverage regions¹¹. Moreover, SNPs can exhibit variability due to random mutations or homoplasy, and their reliability is heavily dependent on sample quality, coverage, and reference genome completeness^12,13.

While comparative genomics using public reference genomes offers an alternative^6,14, it is limited by incomplete references and gaps, leading to missed regions¹³. These challenges highlight the need for more robust methods for accurate species identification, especially in critical fields like agriculture and evolutionary research.

To address these challenges, this study proposes a genomic methodology that leverages species-specific genomic regions to improve classification accuracy. This novel approach provides a robust and reproducible framework for identifying closely related species. As a practical demonstration, we applied this method to distinguish between two genetically similar livestock species: sheep (Ovis aries) and goats (Capra hircus). These species are frequently studied in agricultural and adaptational contexts^3,15,16, making them ideal candidates for testing the effectiveness of this methodology. By incorporating this advanced genomic technique and addressing the limitations of traditional methods, this study aims to contribute to the ongoing efforts to improve species identification accuracy, particularly for closely related species.

Results

To develop a reliable method for distinguishing between closely related species using WGS data, we first analyzed WGS data from 40 samples − 20 sheep (Ovis aries) and 20 goats (Capra hircus) - with confirmed species identity. The samples were sequenced using Illumina NovaSeq 6000, generating an average of 700 M paired-end reads per sample (range: 377-1,081 M reads). Initial quality control showed high-quality sequencing data with consistent read lengths of 150 bp. We first examined standard alignment metrics and taxonomic classification results to evaluate conventional approaches and establish the need for a more specialized method. We then developed and validated a novel approach based on species-specific genomic regions, testing it on both our internal validation set and publicly available data.

Mapping quality metrics

Analysis of mapping quality metrics revealed subtle differences when samples were aligned to species-specific versus non-specific reference genomes (Supplementary Table 1). For goat samples (n = 20), alignment to their native reference (ARS1.2) showed marginally higher alignment rates (98.55–99.45%) compared to alignment against the sheep reference (98.23–99.26%). Similarly, sheep samples (n = 20) aligned to their native reference (ARS-UI_Ramb_v2.0) showed alignment rates of 98.36–99.36% versus 98.13–99.30% when aligned to the goat reference.

For goat samples aligned to their native reference, we observed an average coverage of 35.02X (range: 19.36-53.59X). Interestingly, when these same samples were aligned to the non-native sheep reference, they exhibited slightly higher coverage at 38.90X (range: 21.49-59.49X). Similarly, sheep samples demonstrated comparable coverage patterns, achieving an average of 38.92X (range: 24.39-60.54X) on their native reference and 39.38X (range: 24.75-61.54X) when aligned to the non-native goat reference. The consistently higher coverage observed with the sheep reference genome across all samples suggests that reference genome characteristics, such as shorter total length (ARS-UI_Ramb_v2.0: 2.6Gb vs. ARS1.2: 2.9Gb), may influence coverage more than species-specific factors.

The mismatch rates showed more pronounced but still minor differences: goat samples aligned to their native reference had mismatch rates of 0.63–1.37% versus 2.47–3.11% when aligned to the sheep reference, while sheep samples showed mismatch rates of 0.78–1.29% with their native reference versus 2.44–2.84% with the goat reference. Duplication rates showed slight variations but with overlapping ranges: goat samples showed 1.19–2.77% duplications with their native reference versus 2.84–4.42% with the sheep reference, while sheep samples showed 2.57–3.80% with their native reference versus 3.47–5.92% with the goat reference. These overlapping ranges and subtle differences in mapping metrics demonstrate that conventional alignment statistics alone are insufficient for reliable species discrimination, highlighting the need for a better approach.

Taxonomic classification results

To evaluate conventional taxonomic classification approaches, we analyzed 1,000 paired-end reads (1,000 forward and 1,000 reverse reads) from each sample (n = 40) using Kraken2 with the comprehensive database “k2_nt_20230502” to ensure precise taxonomic classification. Both sheep and goat samples showed predominant assignment to the genus Ovis, followed by Cervus and Bos. As illustrated in Fig. 1, the Sankey diagrams demonstrate substantial overlap in taxonomic assignments between both species.

Fig. 1.

Kraken2 classification of goat and sheep samples. Sankey diagram representation of Kraken2 taxonomic classification results for (A) two goat samples and (B) two sheep samples. The diagrams illustrate the flow of sequence reads from the source samples to their assigned taxonomic classifications. Width of the flow lines corresponds to the proportion of reads assigned to each taxonomic level. This visualization highlights the distribution of reads across different taxonomic categories and demonstrates the challenge in distinguishing between goat and sheep samples using Kraken2’s classification approach. Note the substantial overlap in taxonomic assignments between goat and sheep samples, particularly at higher taxonomic levels.

The boxplot representation (Fig. 2) further quantifies this taxonomic distribution. The distribution patterns of taxonomic assignments were similar between the two species, with substantial overlap in the proportion of reads assigned to different genera. Sheep samples exhibited slightly higher Ovis read counts, and goat samples showed marginally elevated assignments to the Cervus, Bos, and Capra genera.

Fig. 2.

The abundance of genus-level read assignments in goat and sheep samples using Kraken2. The boxplot shows the distribution of genus-level read assignments (n = 1000 paired-end reads) for sheep (n = 20) and goat (n = 20) samples. Both species show similar patterns of taxonomic assignment, with Ovis being the predominant genus in both cases, followed by Cervus and Bos. While sheep samples show higher Ovis read counts and goat samples display slightly elevated Cervus, Bos, and Capra assignments, these differences are insufficient for reliable species discrimination. This overlap in genus-level assignments demonstrates Kraken2’s inability to effectively differentiate between sheep and goat samples.

The high genetic similarity between sheep and goats and conserved sequences shared across related mammalian species explains the ambiguous classification results. Even with our comprehensive reference database, the subtle numeric differences and overlapping taxonomic assignments highlight a key limitation: traditional sequence-based classification methods alone cannot definitively distinguish between these closely related species. This challenge motivated our development of a more qualitative and targeted approach focusing on species-specific genomic regions that can better differentiate between closely related species.

Development of a novel species-specific genomic classification method

Our analysis of standard mapping quality metrics and taxonomic classifications using Kraken2 revealed that conventional approaches are insufficient for reliable species discrimination between closely related species like sheep and goats. This limitation underscores the need for a more targeted and qualitative approach. To address this, we introduce a novel methodology that leverages species-specific genomic regions characterized by their distinctive alignment patterns where they exhibit expected (near-average) coverage depth when samples are mapped to their matching reference genome while displaying minimal to no coverage when aligned to non-matching reference genomes (Fig. 3). We applied and validated this approach using sheep and goat WGS data.

Fig. 3.

Workflow diagram of the species identification methodology. The process consists of two main phases: Training and Testing. In the Training Phase, whole genome sequencing (WGS) data from validated samples of both species are aligned to reference genomes. Species-specific regions are identified and curated as detailed in the workflow. In the Testing Phase, NGS data from an unknown (i.e. unlabeled) sample is aligned to both species’ references, and coverage is calculated in the previously identified species-specific regions. The final species assignment is determined by evaluating coverage patterns: the sample is classified as Species A if coverage is present in A-specific regions but absent in B-specific regions, and vice versa for Species B.

We started with a training phase using 30 WGS samples (15 sheep and 15 goats). Each sample underwent dual alignment to both the sheep (ARS-UI_Ramb_v2.0) and goat (ARS1.2) reference genomes. This dual alignment allowed us to analyze genome-wide coverage patterns and identify regions with species-specific coverage.

By comparing the coverage patterns, we applied a strict selection criterion: regions needed to maintain near-average coverage across all native species samples and less than 2X coverage in all non-native species samples. This process identified 155,800 candidate-specific regions in goats (ranging from 1 to 879 bp) and 1,714,126 candidate-specific regions in sheep (ranging from 1 to 7,928 bp). These regions are listed in Supplementary Tables 2 and 3, respectively.

These candidate regions were manually reviewed, focusing on the longest regions, alignment quality, and genomic context. This curation resulted in the selection of 10 high-confidence, reliable species-specific regions for each species (Table 1). The selected goat-specific regions ranged from 484 to 637 bp in length, while the sheep-specific regions were longer, ranging from 2,178 to 3,884 bp.

Table 1.

Selected species-specific regions. High-confidence species-specific genomic regions identified for species discrimination. The table presents the top 10 curated genomic regions for (A) Goat (Capra hircus) and (B) Sheep (Ovis aries) that show consistent species-specific coverage patterns. Each region is characterized by robust coverage when samples are aligned to their native reference genome and negligible coverage (< 1X) in non-native alignments. Regions are listed with their genomic coordinates using the ARS1.2 reference for goat and ARS-UI_Ramb_v2.0 reference for sheep. Region ID is a unique region identifier (GR1-GR10 for goat regions, SR1-SR10 for sheep regions).

Region ID	Contig ID	Chromosome	Start	End	Length (bp)
(A) Goat
GR1	NC_030811.1	chr4	9,189,686	9,190,241	555
GR2	NC_030815.1	chr8	1,639,607	1,640,124	517
GR3	NC_030815.1	chr8	89,033,445	89,033,929	484
GR4	NC_030817.1	chr10	14,363,141	14,363,768	627
GR5	NC_030818.1	chr11	19,059,679	19,060,308	629
GR6	NC_030819.1	chr12	78,876,484	78,877,107	623
GR7	NC_030822.1	chr15	27,911,796	27,912,433	637
GR8	NC_030825.1	chr18	9,518,750	9,519,236	486
GR9	NC_030827.1	chr20	31,744,884	31,745,471	587
GR10	NC_030833.1	chr26	23,762,652	23,763,176	524
(B) Sheep
SR1	NC_056055.1	chr2	72,752,582	72,754,965	2,383
SR2	NC_056058.1	chr5	82,801,325	82,804,058	2,733
SR3	NC_056060.1	chr7	53,785,302	53,787,489	2,187
SR4	NC_056062.1	chr9	84,555,330	84,559,214	3,884
SR5	NC_056063.1	chr10	41,611,380	41,613,958	2,578
SR6	NC_056068.1	chr15	14,655,024	14,658,653	3,629
SR7	NC_056068.1	chr15	34,033,704	34,036,921	3,217
SR8	NC_056070.1	chr17	35,040,542	35,043,956	3,414
SR9	NC_056075.1	chr22	21,693,259	21,695,437	2,178
SR10	NC_056076.1	chr23	29,548,956	29,551,728	2,772

To evaluate if the selected identified regions can serve as reliable markers for species identification in WGS data analysis. The testing phase involves a simple step-by-step procedure (Fig. 3): (1) Align the unlabeled sample to both the sheep and goat reference genomes; (2) Calculate the coverage of the selected sheep-specific regions after aligning to the sheep genome; (3) Calculate the coverage of the selected goat-specific regions after aligning to the goat genome; and, (4) If coverage is nearly zero in the sheep-specific regions but near average in the goat-specific regions, the sample is classified as a goat. Conversely, if the coverage pattern is reversed, the sample is classified as a sheep.

This approach achieved 100% accuracy within the analyzed validation datasets comprising 14 independent samples from multiple sources and sequenced on different Illumina platforms (Table 2). The selected regions showed remarkable consistency, with all native species samples displaying expected coverage while all non-native species exhibited low or absent coverage (Fig. 4). While these coverage patterns remained robust across varying sequencing depths, bisulfite-converted WGS samples maintained readily distinguishable patterns, though with slightly reduced contrast (Table 2). Consistent performance across diverse sample sources and sequencing platforms demonstrated both the reliability and broad applicability of this method, consistently achieving unambiguous classification.

Table 2.

Coverage analysis of species-specific regions across validation samples. The table shows coverage depth for each species-specific region (rows) across different samples (columns). Goat samples (G1–G7) include five internal validation samples (G1–G5) and two public samples: G6 (ERR3988607) and G7 (ERR8529066). Similarly, sheep samples (Sh1–Sh7) include five internal validation samples (Sh1–Sh5) and two public samples: Sh6 (ERR10368013) and Sh7 (ERR10368017). Region IDs (GR1–GR10 for goat-specific regions and SR1–SR10 for sheep-specific regions) correspond to the genomic coordinates detailed in Table 1. Values represent average coverage depth. Native species show robust coverage in their corresponding species-specific regions, while non-native species exhibit minimal to no coverage in these same regions. This clear dichotomy in coverage patterns provides reliable discrimination between species. *Mean sequencing depth using the matched reference genome of the sample.

Region/sample	G1	G2	G3	G4	G5	G6	G7	Sh1	Sh2	Sh3	Sh4	Sh5	Sh6	Sh7
WGS coverage*	36×	35×	31×	35×	28×	11×	19×	42×	36×	36×	57×	35×	23×	18×
GR1	27.2	30.2	26.7	27.6	24.2	8.3	12.5	0.0	0.0	0.3	0.0	0.6	0.1	0.0
GR2	40.5	37.4	37.1	43.9	39.4	8.7	12.6	0.0	0.0	0.0	0.0	0.6	0.9	0.4
GR3	31.7	40.1	27.9	35.2	27.6	10.0	14.4	0.3	0.0	0.0	0.6	0.0	0.0	0.2
GR4	38.4	35.1	27.1	38.7	22.9	12.4	8.7	0.5	0.0	0.0	0.0	0.0	0.0	0.1
GR5	36.6	41.2	35.3	32.5	27.7	9.3	11.0	0.0	0.0	0.0	0.4	0.0	0.1	0.0
GR6	40.9	36.9	37.1	48.1	31.8	7.5	12.6	0.0	0.0	0.0	0.1	0.0	2.3	1.8
GR7	39.0	33.7	31.0	26.8	32.5	11.4	14.4	0.0	0.0	0.0	0.0	0.0	1.1	0.4
GR8	26.3	28.7	24.5	27.1	25.3	3.7	17.0	0.0	0.0	0.0	0.0	0.1	0.0	0.0
GR9	32.9	34.3	24.1	32.0	26.6	10.1	13.9	0.0	0.0	0.0	0.2	0.0	0.1	0.1
GR10	43.6	37.6	30.6	39.3	31.5	9.4	11.3	0.0	0.0	0.0	0.2	0.0	1.6	1.7
SR1	0.0	0.0	0.1	0.0	0.0	0.0	0.0	34.0	32.2	25.7	50.9	27.8	7.5	6.6
SR2	0.0	0.0	0.0	0.1	0.0	0.0	0.0	38.9	33.0	35.8	54.3	29.5	10.5	7.7
SR3	0.0	0.0	0.0	0.0	0.0	0.0	0.0	41.3	33.2	36.7	60.3	34.2	12.6	10.8
SR4	0.0	0.0	0.0	0.1	0.1	0.0	0.0	39.6	35.7	37.8	54.9	31.8	11.8	8.4
SR5	0.0	0.0	0.0	0.0	0.0	0.0	0.0	41.3	30.9	33.9	56.2	34.0	11.9	7.5
SR6	0.0	0.0	0.0	0.0	0.0	0.0	0.1	42.1	33.2	34.2	55.8	34.1	6.7	6.7
SR7	0.0	0.0	0.1	0.0	0.0	0.0	0.0	37.4	31.9	33.5	50.5	31.7	7.1	6.6
SR8	0.3	0.1	0.3	0.1	0.0	0.0	0.1	41.1	31.5	36.5	53.4	32.2	8.1	7.5
SR9	0.1	0.0	0.2	0.0	0.0	0.0	0.0	40.7	33.6	32.8	47.2	29.1	5.7	5.4
SR10	0.0	0.1	0.0	0.1	0.0	0.0	0.2	44.5	32.6	28.7	51.7	30.3	9.0	6.1

Fig. 4.

Visualization of species-specific genomic regions using IGV (Integrative Genomics Viewer). (A) Coverage patterns in a sheep-specific region (chromosome 23: 29,548,756 − 29,551,928) showing clear differentiation between species. The top three tracks display consistent coverage in validated sheep samples, while the bottom three tracks from goat samples show no coverage when aligned to the sheep reference genome. (B) Coverage patterns in a goat-specific region (chromosome 8: 89,033,215 − 89,034,298) demonstrating similar species discrimination. The top three tracks show robust coverage in validated goat samples, while the bottom three tracks from sheep samples exhibit no coverage when aligned to the goat reference genome. These contrasting coverage patterns demonstrate the utility of these regions for species discrimination.

Discussion

Our analysis of standard mapping quality metrics and taxonomic classifications using Kraken2 revealed that conventional approaches are insufficient for reliable species discrimination between closely related species like sheep and goats. In this study, we introduced a robust, targeted, and qualitative genomic approach for accurately identifying species, particularly useful for closely related species like sheep and goats. Our method is designed specifically for analyzing already generated WGS data rather than applying it at the DNA sample level or during the steps leading up to sequencing. The achieved accuracy (i.e. 100%) in distinguishing between our analyzed validation set of 14 independent sheep and goat WGS samples highlights the potential of this method for broader applications in biological and agricultural research, which should be explored in future research.

A key strength of the proposed approach is its empirical nature, which bypasses the limitations that face comparative genomic studies, such as incomplete or biased reference sequences^17–20. By employing WGS data, we gain a comprehensive representation of the genetic material present in the samples. If a region in the reference genome lacks coverage in the WGS data, it indicates that this region may be specific to the reference species. This enables informed distinctions based on the presence or absence of these unique genomic sequences.

One of the key outputs of our method is a list of species-specific genomic regions. This list is crucial for our identification process and opens exciting paths for further research. Scientists can use these regions to investigate their biological importance, potentially uncovering new insights into species-specific traits or adaptations. Moreover, these identified regions can serve as a foundation for developing more targeted molecular methods, such as PCR-based assays or DNA probes^20–24, which could offer quicker and more cost-effective species identification tools for field use.

The number of regions selected represents a pragmatic design choice rather than a strict methodological requirement. In principle, a single highly reliable region is sufficient for correct species classification based on coverage patterns; however, reliance on a single region may increase sensitivity to rare technical artifacts or biological variation across breeds or populations. To improve robustness and reduce the risk of misclassification, we therefore selected multiple independent regions. In this study, ten regions per species were chosen as a balanced compromise between redundancy and computational simplicity.

While our study concentrated on binary classification (e.g. sheep versus goats), the methodology shows promise for extending to multiclass problems^25,26, potentially allowing us to differentiate among multiple close-related species using their WGS data. However, scaling up this approach does come with challenges, particularly in terms of the computational resources and time required to align and classify large NGS datasets. One potential optimization could be implementing a checkpoint system to verify alignment matches with species-specific regions - if these regions lack coverage, the sample could then be realigned specifically against closely related species. Future research should focus on exploring ways to reduce these resource demands while maintaining high accuracy.

Sample mislabeling due to human error poses a significant challenge in research^6,27–29, particularly in dynamic environments like zoos, veterinary clinics, and field sites. When closely related species are housed together, the risk of misidentification increases, potentially compromising valuable data. Our genomic-based approach offers a robust solution to this challenge, providing a reliable method to verify and correct sample metadata (Fig. 5). This is especially valuable when visual differences between species are subtle, helping maintain data integrity across biological research disciplines.

Fig. 5.

This workflow illustrates a method for distinguishing between two species using whole-genome sequencing data. Reads from both species are aligned to each other’s reference genomes to find regions uniquely covered by one species. These species-specific regions form the basis for classifying unknown samples, with coverage patterns enabling accurate and reliable identification.

We acknowledge the burden of the current classification method, which requires double computational resources for alignment and storage, as both reference genomes must be aligned to the sample data for accurate classification. Additionally, our method cannot provide species classification directly from raw FASTQ files; alignment is a necessary step before any identification can be made. As the approach relies exclusively on genomic coverage patterns, intra-species phenotypic variation that is not associated with major genomic differences is not expected to influence classification performance; however, evaluation across diverse breeds and populations remains an important direction for future work. Addressing these limitations represents a significant area for improvement in future research, where optimizing alignment processes and reducing storage requirements could enhance the efficiency and practicality of our approach.

To support researchers in adopting our method, we have made a comprehensive set of resources available in our GitHub repository, specifically tailored for working with NGS data files. This includes illustrative example code, a list of candidate regions, top-curated regions critical to our analysis, and test results that validate the effectiveness of our method. These resources make it easy for researchers to adopt and adapt our approach in their studies of existing NGS datasets, fostering seamless integration into various research projects and enhancing reproducibility in species identification.

This study not only provides a practical solution for identifying sheep and goats from WGS data but also lays the groundwork for addressing similar challenges in other species. The species-specific regions we have identified open new research possibilities and potential molecular tools. As genomic technologies and NGS data analysis techniques continue to evolve³⁰, integrating this methodology into automated bioinformatics pipelines could further enhance its utility. Ultimately, our work contributes to broader efforts in biodiversity assessment and conservation genetics, all while focusing on the analysis of post-sequencing data and providing valuable resources for future research and tool development.

Materials and methods

Sample collection

Blood samples were ethically collected from sheep (Ovis aries, n = 20) and goats (Capra hircus, n = 20) from different locations in Abu Dhabi and Al Ain in the United Arab Emirates (UAE) and stored in refrigerated containers to ensure proper preservation during transportation to the laboratory. The ethical approval was obtained from the Animal Research Oversight Committee of Khalifa University of Science and Technology (Protocol #: A20-003). All procedures were conducted in accordance with institutional, national, and international guidelines for the care and use of animals in research such as the ARRIVE guidelines (https://arriveguidelines.org).

Data processing and alignment

Genomic DNA was extracted from the collected blood samples and subjected to PCR-free library preparation using the Illumina DNA Prep kit (i.e. Nextera DNA Flex). The prepared libraries were then sequenced on the Illumina NovaSeq 6000 platform, generating 150 bp paired-end reads. Sequencing data underwent quality control and adapter trimming using standard Illumina pipelines. Alignment to reference genomes was performed using the BWA-MEM algorithm through the Sentieon v202112.07 software (Sentieon; https://www.sentieon.com). Mapping quality measurements were generated by Sentieon software using Picard tools. For sheep, the ARS-UI_Ramb_v2.0 reference genome was utilized, while for goats, the ARS1.2 reference was utilized. Coverage in aligned BAM files was determined using the “bedtools genomecov” command.

Taxonomic classification using Kraken2

To evaluate the effectiveness of conventional taxonomic classification methods, we employed Kraken2 for species identification⁸. A subset of 1000 paired-end reads (1000 forward and 1000 reverse reads) was randomly selected from each sample for analysis. The comprehensive database “k2_nt_20230502” was used to ensure precise taxonomic classification. Kraken2 (version 2.1.3) was run with the following parameters: kraken2 --quick --memory-mapping --minimum-base-quality 10 --db k2_nt_20230502. Kraken2 outputs were visualized using Pavian to explore classification results³¹. The subset of 1000 paired-end reads was used primarily for computational efficiency given the large Kraken2 database, and additional tests using 50,000 reads produced highly similar results, confirming that the smaller subset was sufficient for the analysis (Supplementary Fig. 1).

Sample selection for validation

To validate our method, we used both internal and publicly available data. We randomly selected 10 samples, 5 goat and 5 sheep samples, from our cohort that were not used in the training process. This internal validation set allowed us to assess the method’s performance on data collected under similar conditions to our training set. We also extended our validation dataset to the publicly available European Nucleotide Archive (ENA, www.ebi.ac.uk). For goats, we selected two samples: ERR3988607³², sequenced on the Illumina HiSeq 2000 platform, and ERR8529066³², sequenced on the Illumina NovaSeq 6000 platform. For sheep, we chose ERR10368013 and ERR10368017³³, both sequenced on the Illumina HiSeq 3000 platform. All samples were paired-end WGS datasets, except for sheep samples ERR10368013 and ERR10368017, which were bisulfite-converted WGS samples. All samples have a read length of 150 bp except ERR3988607, which has 100 bp. This diverse set of public samples allowed us to evaluate our method’s performance across various sequencing technologies and sample sources, enhancing the generalizability of our findings.

Implementation resources

To facilitate the application of our species identification method, we have developed a comprehensive toolkit that is available in our GitHub repository (https://github.com/BTC-Lab/Goat_Sheep_specific_regions). This includes illustrative example code that demonstrates the practical implementation of our approach, a comprehensive list of candidate regions, and selected top-curated regions that were critical to our analysis. Additionally, the repository contains BAM file screenshots documenting the coverage in key genomic regions and test results that validate the method’s effectiveness. While the methodology is designed for closely related species, we further verified the selected regions using horse and camel WGS alignments, confirming the absence of meaningful coverage in these non-target species; representative alignment screenshots are provided in the GitHub repository. The core of the implementation is a coverage calculation script that can be executed using a simple command: `sh calculate_coverage.sh ${regions_bed} ${bamFile}`. This script enables straightforward calculation of coverage for specified genomic regions in a given BAM file, making our method accessible and adaptable to various research contexts.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ARS1.2

Goat reference genome assembly

ARS-UI_Ramb_v2.0

sheep reference genome assembly

BAM

Binary Alignment/Map format

BLAST

Basic Local Alignment Search Tool

bp

Base pairs

BWA-MEM

Burrows-Wheeler Aligner Maximal Exact Matches

ENA

European Nucleotide Archive

IGV

Integrative Genomics Viewer

NGS

Next-Generation Sequencing

NCBI

National Center for Biotechnology Information

PCR

Polymerase Chain Reaction

SNP

Single Nucleotide Polymorphism

SRA

Sequence Read Archive

UAE

United Arab Emirates

WGS

Whole Genome Sequencing

Author contributions

**Nour al dain Marzouka: ** Conceptualization; Data Curation; Methodology; Analysis; And Writing – Original Draft.**Amira Al-Aamri: ** Data Processing; Data Curation; Methodology; Analysis; Visualization; Writing – Review and Editing.**Fatima Alshamsi: ** Resources; Visualization; Writing – Review and Editing.**Mariam Khalili: ** Resources; Writing – Review and Editing.**Sarah El Hajj Chehadeh** : Resources; Writing – Review and Editing.**Meera S. Mohamed: ** Project Administration; Writing – Review and Editing.**Yassir Mohammed Eltahir: ** Project Administration; Writing – Review and Editing.**Rafeek Koliyan: ** Project Administration; Writing – Review and Editing.**Mohamed Moustafa Abdulhalim: ** Project Administration; Writing – Review and Editing.**Assem Attia: ** Project Administration; Writing – Review and Editing.**Mira Mousa: ** Project Administration; Writing – Review and Editing.**Guan Tay: ** Project Administration; Writing – Review and Editing.**Habiba Alsafar: ** Supervision; Project Administration; Writing – Review and Editing.

Funding

This research did not receive any external funding.

Data availability

Sequences have been submitted to the NCBI Sequence Read Archive (SRA) database under the BioProject accession number PRJNA1120375 ( [https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1120375](https:/www.ncbi.nlm.nih.gov/bioproject/PRJNA1120375) ) and SRA Accession: SRP511937.

Declarations

Competing interests

The authors declare no competing interests.

Ethics statement

The collection of blood samples and all associated procedures were approved by the Animal Research Oversight Committee of Khalifa University of Science and Technology (Protocol #: A20-003). All animal experiments were conducted in compliance with institutional, national, and international guidelines. The study is reported in accordance with the ARRIVE guidelines.

Declaration of AI-assisted technologies in the writing process

In this manuscript, AI-assisted technologies were utilized exclusively to enhance the language clarity and readability.