A pig structural variation reference panel based on graphical pan-genome and construction method and application thereof

Abstract

The application discloses a pig structural variation reference panel based on a graphic pan-genome and a construction method and application thereof, and belongs to the field of biological molecular breeding and genomic information analysis.The method is based on large-scale pig population high-depth whole genome sequencing data to construct a graphic pan-genome, realize unified discovery, accurate positioning and standardized coding of structural variations, and construct a comprehensive reference panel containing SNPs, INDELs and SVs.The constructed reference panel integrates 6449 pig whole genome data of 154 breeds, contains about 5030 million SNPs, 5.8 million INDELs and 253 thousand SVs, and is one of the largest and most genetically diverse pig reference panels at present.Further, the panel is used for SNP and SV joint filling of low-depth resequencing or SNP chip data, so that high-precision genotyping of structural variations of target individuals is realized, and important technical support is provided for pig genetic breeding and complex trait analysis.

Description

Technical Field

[0001] This invention belongs to the field of biomolecular breeding and genome information analysis technology, specifically relating to a pig structural variation reference panel based on a graphical pangenome, its construction method, and its application. Background Technology

[0002] Pigs are one of the most important livestock species globally, occupying a central position in large-scale farming and molecular breeding systems. With the development of modern breeding technologies, large-scale breeding companies typically use imported breeds or their hybrid offspring as the core breeding population, placing higher demands on the accuracy, consistency, and scalability of genomic information. However, in large-scale breeding practices, obtaining comprehensive and accurate genomic variation information while controlling costs remains a significant technical bottleneck restricting the efficiency of molecular breeding.

[0003] Current pig genome research and genotyping analysis typically rely on a single linear reference genome for sequencing data alignment and variant detection. The widely used pig reference genome, Sscrofa11.1, is derived from individual European Duroc pigs. While it represents the genetic background of commercial pig populations to some extent, a significant number of genetic variations remain outside the linear reference genome across different introduced breeds, crossbred populations, and long-term selectively bred lines. When using a single linear reference genome to align sequencing data from pig populations with diverse genetic backgrounds, reference bias can easily arise, leading to alignment failures or errors in some sequences, thus affecting the completeness and accuracy of genotyping.

[0004] The aforementioned reference bias problem is particularly prominent in the detection of structural variations. Structural variations include complex types such as insertions, deletions, inversions, and copy number variations, a significant portion of which involve genomic fragments missing from the reference genome or non-reference sequences. Because a linear reference genome cannot fully represent these variations, existing techniques are prone to missed detections, false positives, or inconsistencies in allelic status determination during structural variation detection and genotyping, thus affecting subsequent genetic analysis results.

[0005] Structural variations play a crucial role in regulating gene expression, influencing growth and development, environmental adaptability, and the formation of important economic traits. Accurate detection and genotyping of structural variations are essential for population genetic structure analysis, genome-wide association studies, and genomic selection breeding. However, existing methods for detecting and genotyping structural variations typically rely on high-depth whole-genome sequencing data, which incurs high sequencing and data processing costs, making long-term application in breeding populations of tens of thousands difficult.

[0006] In actual production and breeding processes, low-depth whole-genome sequencing and SNP chip data are widely used due to their low cost and high throughput. However, under the above data conditions, existing technologies mainly focus on the analysis of single nucleotide polymorphisms and insertion / deletion variations, making it difficult to achieve reliable typing of structural variations. In particular, there is a lack of a technical solution that can uniformly define and accurately fill in structural variations in commercial pig herds of multiple populations and breeds.

[0007] In recent years, haplotype reference templates and genotype imputation methods have made some progress in improving the utilization efficiency of low-density genotype data. However, most existing reference templates are based on linear reference genomes and mainly consist of single nucleotide polymorphisms and insertion / deletion variations, which are insufficient in terms of coverage and consistent definition of structural variations. This makes it difficult to fundamentally solve the problems of reference bias and the difficulty in genotyping structural variations under low-cost conditions. Therefore, there is an urgent need for a new low-cost genotyping technology that can integrate genetic variation information from multiple populations, uniformly represent structural variations, and is applicable to low-depth sequencing and microarray data. Summary of the Invention

[0008] Technical Problem to be Solved: To address the aforementioned technical problems, the purpose of this invention is to provide a graphical pan-genome-based reference panel for pig structural variation, its construction method, and its applications. By integrating medium- to high-depth whole-genome sequencing data from multiple pig populations, a graphical pan-genome model containing non-reference sequences is constructed. Structural variations are discovered, located, merged, and allelically encoded within a unified reference framework, thereby constructing a comprehensive reference panel containing SNPs, InDels, and SVs. Based on this reference panel, this invention can perform joint SNP, InDel, and SV imputation and genotyping on low-depth whole-genome sequencing data or SNP microarray data, enabling target individuals to obtain reliable structural variation genotype information without requiring high-depth sequencing. This method significantly reduces the cost of large-scale pig herd structural variation detection and genotyping while ensuring consistency in structural variation definitions and genotyping accuracy. It is suitable for applications such as breed identification, population genetic structure analysis, genome selection breeding, and genome-wide association studies.

[0009] Technical solution: A pig structural variation reference panel based on graphical pangenome, namely the PGIP reference panel, contains 6449 pig individuals from 154 breeds or populations, integrating 50.3 million single nucleotide polymorphisms (SNPs), 5.8 million insertion / deletion variants (InDels), and 253,286 structural variants (SVs); the structural variants (SVs) cover coding regions, intron regions, and intergenic regions, and can be divided into overlapping windows with buffers according to chromosomes to establish a reference panel index.

[0010] Preferably, the pig individuals include local breeds, commercial breeds, and their bred or hybrid populations.

[0011] The above method for constructing a pig structural variation reference panel based on a graphical pan-genome includes the following steps:

[0012] S1. Acquire high-depth whole-genome sequencing data of pigs from multiple breeds or populations, perform quality control, reference alignment, sorting, deduplication and variant detection on the sequencing reads, and obtain standardized sequencing data for single nucleotide polymorphisms (SNPs), insertion / deletion variants (InDel), and structural variant (SV) analysis.

[0013] S2. Based on sequencing data and the pig linear reference genome Sscrofa11.1, a graphical pan-genome model containing non-reference sequences was constructed, integrating sequences missing from the linear reference genome, insertion sequences, and representative haplotype pathways from different pig individuals or populations into a unified graph structure;

[0014] S3. In the graphical pangenome model, structural variations are uniformly discovered, merged, located, and allelically encoded to obtain a standardized definition system of structural variations that includes insertions, deletions, inversions, duplications, copy number variations, and presence / deletion variations;

[0015] S4. Based on the standardized structural variation definition system, perform consistent structural variation genotyping on multiple pig individuals, and integrate it with SNP and InDel genotype data to construct a pig structural variation reference panel containing SNP, InDel, and SV.

[0016] Preferably, the sequencing data in step S1 is aligned to the Sscrofa11.1 reference genome after quality control, and SNPs and InDels are jointly detected, filtered, and normalized.

[0017] Preferably, the graphical pangenome model described in step S2 reduces reference bias by introducing non-reference sequences that are missing in the linear reference genome.

[0018] Preferably, the structural variations described in step S3 are merged according to chromosome, breakpoint position, interval overlap, insertion sequence similarity, and allele path consistency, and each structural variation is assigned a unique number and a unified allele code.

[0019] A method for structural variation genotyping includes the following steps:

[0020] (1) Obtain low-depth whole genome sequencing data or SNP chip data of the target individual and match them with the reference panel in terms of loci, coordinates and allele orientation;

[0021] (2) Based on the reference panel and unified structural variation definition and coding system described in claim 1 or 2, SNP, InDel and SV joint imputation analysis is performed on the target individual to obtain the structural variation genotype results of the target individual.

[0022] Preferably, the sequencing depth of the low-depth whole genome sequencing data in step (1) is 0.5×~2×; the SNP chip data is low-density or medium-density SNP chip data; the joint imputation analysis uses haplotype linkage disequilibrium information, genotype likelihood or chip site information in the reference panel to simultaneously imput the SNPs, InDel and SV of the target individual, and filters the results according to INFO or equivalent imputation quality indicators.

[0023] The above-mentioned graphical pan-genome-based reference panel for pig structural variation has been applied to pig genetic improvement and molecular breeding.

[0024] The above-mentioned structural variation gene typing method is applied in the genetic improvement and molecular breeding of pigs.

[0025] Beneficial effects:

[0026] 1. This invention provides a Porcine Graph-based Integrated Panel (PGIP) and genotyping system for structural variations based on a graph pan-genome. The system comprehensively considers sample size, population diversity, and variation type coverage during the construction of structural variations. Through a unified strategy for discovering, locating, and encoding structural variations, it ensures that structural variations from the same source in different individuals and populations have a unified definition and representation, thereby improving the coverage integrity and genotyping consistency of structural variations across the entire genome and avoiding the problems of missed detections and misclassifications caused by reference bias under traditional linear reference genome conditions.

[0027] 2. This invention provides a joint imputation strategy based on a structural variation reference panel, which can simultaneously imput single nucleotide polymorphisms and structural variations using low-depth whole-genome sequencing data or SNP microarray data as input, and obtain reliable structural variation genotype information without the need for high-depth sequencing. The imputed genome data can meet the needs of subsequent research and applications such as variety identification, population genetic structure analysis, genome selection breeding, and genome-wide association analysis in terms of accuracy and completeness.

[0028] 3. This invention can significantly reduce detection costs while ensuring the accuracy and stability of structural variant genotyping. It breaks through the dependence of structural variant analysis on high-depth whole-genome sequencing, making structural variant genotyping technology applicable to genetic assessment and molecular breeding scenarios in large-scale pig herds. This improves the availability and promotion value of structural variant information in actual breeding production and has a higher cost-performance ratio.

[0029] 4. This invention constructs a unified structural variation reference framework based on a graphical pan-genome. Compared with traditional structural variation detection methods based on linear reference genomes, it can effectively reduce the reference bias generated during sequencing data alignment and variation identification in pig populations with different genetic backgrounds. This not only improves the detection rate and typing accuracy of structural variations, but also provides stable and reliable basic data support for subsequent association analysis of structural variations with gene expression regulation, environmental adaptability, and important economic traits. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the construction of a reference panel for pig structural variation based on a graphical pangenome and a genotyping method.

[0031] Figure 2 This section presents the basic characteristics (MAF, length distribution, proportion) of SVs in the porcine structural variation panel, as well as population genetic analysis (PCA, NJ-tree, Admixture). A: Minimum allele frequency density distribution map for different structural variation types; B: Length analysis for different structural variation types (insertion, deletion, all); C: Bar plot of structural variation proportions under different MAFs; D: Principal component analysis (PCA) distribution map based on structural variation genotypes; E-F: Neighbor-Joining tree (NJ-tree) constructed based on structural variations; G: Population genetic structure analysis diagram based on structural variations (Structure / Admixture analysis results).

[0032] Figure 3 The graphs show the changes in fill-in accuracy for different varieties under different sequencing conditions. A: Changes in fill-in accuracy for structural variants under low-depth sequencing conditions (0.5×, 0.8×, 1×) and high-depth WGS real set conditions for different varieties. B: Changes in fill-in accuracy for major and minor structural variants under low-depth sequencing conditions (0.5×, 0.8×, 1×). C: Changes in fill-in accuracy for structural variants within different allele frequency ranges under low-depth sequencing conditions (0.5×, 0.8×, 1×) for different varieties. Detailed Implementation

[0033] The present invention will be further described below with reference to embodiments. These embodiments are illustrative of the invention and are not intended to limit the scope of protection of the invention. The embodiments are used to illustrate specific implementations of the invention and are not intended to limit the scope of protection of the invention. Without departing from the concept of the invention, those skilled in the art can make equivalent adjustments to the software version, parameter thresholds, and window size according to the sequencing platform, sample size, pig herd origin, and computing environment.

[0034] Example 1

[0035] This embodiment describes the construction of a Porcine Graph-based Integrated Panel (PGIP) for pig structural variation. The specific method includes the following steps:

[0036] S1. Acquire whole-genome sequencing data of pigs from multiple breeds and perform basic variant detection. Collect whole-genome sequencing data of 6449 pig individuals from 154 breeds or populations, including local breeds, commercial breeds, and their breeding or hybrid populations. Assign a unique number to each sample, recording the breed name, population origin, sequencing platform, sequencing depth, data batch, file path, and subsequent analysis grouping, forming a sample information table. If multiple sequencing batches exist for the same sample, first merge homologous FASTQ data according to sample number, and retain batch information for subsequent anomaly sample tracking.

[0037] Quality control was performed on the raw FASTQ data. Specifically, fastp or equivalent software was used to remove adapter sequences, low-quality bases, and reads containing excessive N bases; consecutive N sequences at the beginning and end of reads were removed; reads with an average base quality lower than Q16 or an effective length shorter than 36 bp were filtered; the number of reads before and after filtering, the Q20 / Q30 ratio, GC content, and retention rate were output for each sample. Samples with significantly lower effective read counts, abnormal GC content, or significantly insufficient sequencing depth after filtering were marked as samples requiring further review and were not included in the main reference panel construction process or were only used for auxiliary analysis.

[0038] High-quality reads after quality control were aligned to the porcine linear reference genome Sscrofa11.1. BWA-MEM or equivalent short-read alignment software could be used for alignment to generate a SAM file. Subsequently, samtools or equivalent tools were used to convert, sort, and index the reads, and Picard or equivalent tools were used to label PCR repeats. The alignment results were statistically analyzed for average sequencing depth, coverage, repetition rate, insert length distribution, and alignment rate. Unaligned reads, low-alignment-quality reads, and obviously abnormal repetitive reads were filtered out to obtain a sorted and indexed standardized BAM file. This BAM file also serves as the basic input for SNP / InDel detection, structural variant genotyping, and subsequent low-depth simulation evaluation.

[0039] SNPs and InDels were detected based on standardized BAM files. Specifically, a single-sample gVCF file was generated for each sample using GATK HaplotypeCaller or similar software, and then joint genotyping was performed on all samples to obtain multi-sample VCF files. SNP sites were filtered using criteria such as QD < 2.0, ReadPosRankSum < −8.0, FS > 60.0, MQ < 40.0, or MQRankSum < −12.5. InDel sites were filtered based on binding site quality, sequencing depth, deletion rate, allele balance, and proximity to complex sequences. The filtered VCFs were then left-aligned, multi-allele sites were split, duplicate sites were removed, and reference / substitute allele orientations were corrected to ensure consistency across all samples within the same chromosome, at the same location, and with the same allele definition.

[0040] The outputs of step S1 include: a quality-controlled FASTQ file, a normalized BAM file, a multi-sample SNP / InDel VCF file, a sample quality control table, and a locus quality control table. The multi-sample SNP / InDel VCF serves as the SNP / InDel portion of the subsequent PGIP reference panel, while the normalized BAM file serves as a source of read evidence for graphical pangenome construction, SV discovery, and SV genotyping.

[0041] S2. Construct a graphical pan-genome model including non-reference sequences. Using Sscrofa11.1 as the backbone reference path, collect candidate non-reference sequences from various pig breeds that cannot be fully represented by the linear reference genome. Candidate non-reference sequences can come from assembled sequences of representative individuals, incompletely aligned reads, soft-splitting read splicing results, inserted SV sequences, and existing non-reference sequence sets. Redundancy removal, low-complexity filtering, and length filtering are performed on the candidate sequences, retaining sequences that are locatable, can be supported by reads, and have clear breakpoints or alternative paths.

[0042] Candidate non-reference sequences are aligned with Sscrofa11.1 to determine their chromosome location, insertion breakpoint, substitution region, or paraphyletic path. Non-reference sequences that can be uniquely located are used as non-reference nodes or substitution paths in the graph structure. For multiple candidate sequences that are highly similar at the same breakpoint, representative sequences are retained and their supporting samples are recorded. Sequences that cannot be located, are overly repetitive, or lack sufficient supporting evidence are temporarily excluded from the main graph structure. Reference paths, non-reference insertion paths, deletion paths, and representative haplotype paths are retained during graph construction, allowing the same genomic region to simultaneously represent reference and non-reference alleles.

[0043] After completing the graphical pangenome model, an index was created for the graph structure, and the correspondence between each non-reference path and the Sscrofa11.1 backbone coordinates was recorded. This correspondence includes chromosome number, reference start and end coordinates, breakpoint location, non-reference sequence length, path number, and supporting evidence. Through this step, subsequent structural variations no longer rely solely on linear reference genome coordinates but are represented in the graph structure using uniform paths and breakpoints, thereby reducing missed detections and misclassifications due to reference bias in pig populations with different genetic backgrounds.

[0044] The outputs of step S2 include: a graphical pangenome index file, a non-reference sequence path table, a reference coordinate to graph path mapping table, and a candidate non-reference sequence quality control table. These files together form the reference framework for subsequent unified SV discovery, merging, and encoding.

[0045] S3. Unified discovery, merging, localization, and encoding of structural variations in the graphical pangenome. The standardized BAM file or post-quality-controlled reads obtained in step S1 are projected into the graphical pangenome model constructed in step S2. Structural variations are identified based on various pieces of evidence, including: cross-breakpoint read support, paired read anomalies, soft-splitting sequences, local read depth changes, non-reference path support, and allelic path coverage. Identified SV types include insertions, deletions, inversions, duplications, copy number variations, and variations containing the presence / deletion of non-reference sequences.

[0046] Candidate SVs obtained from different samples, different test batches, or different evidence sources are merged and standardized. For deletions, duplications, and copy number variations, homologous events are merged according to chromosome, start and end coordinates, breakpoint distance, and interval overlap ratio. For insertion variations, homologous events are merged according to insertion breakpoint distance, insertion sequence similarity, insertion length, and graph path consistency. For directional events such as inversions, confirmation is based on breakpoint direction and paired read direction. During merging, if two candidate events have close breakpoints and support the same non-reference path or the same interval variation, they are merged into the same SV; if the breakpoints are close but the allelic sequences or paths are significantly different, they are retained as different allelic events.

[0047] A standardized definition was established for each merged SV. Each SV recorded a unique ID, chromosome, reference start and end coordinates, variant type, SV length, reference allele, substitute allele, corresponding graph path, number of supporting samples, number of supporting reads, and initial allelic frequency. For biallelic SVs, 0 represented the reference allele and 1 represented the substitute allele; multiallelic or copy number-related SVs were first split or converted into allelic codes suitable for haplotype filling before being included in panel construction. This standardized definition system ensures that SVs from the same source have the same coordinates, IDs, and allelic meanings in different individuals and populations.

[0048] The output of step S3 is an SV definition file and a set of SV candidate sites. This set is not the final reference panel, but rather a site template for subsequent consistent SV genotyping of 6449 panel samples.

[0049] S4. Perform consistent SV genotyping on all individuals within the panel and construct a PGIP reference panel. Using the SV definition file obtained in step S3 as a fixed template, perform SV genotyping on each of the 6449 pig individuals. For each sample and each SV locus, statistically analyze the readings supporting the reference allele and the alternative allele, graph path coverage, breakpoint support, and local read depth changes, and determine the genotype accordingly. Biallelic SV genotypes are encoded as 0 / 0, 0 / 1, or 1 / 1; when supporting evidence is insufficient, genotype quality is too low, local depth is abnormal, or allele evidence is contradictory, the genotype of that sample at that SV locus is marked as missing.

[0050] After completing single-sample SV genotyping, the SV genotypes of all samples are merged into a multi-sample SV matrix, and site-level quality control is performed. Quality control includes: removing SV sites with excessively high deletion rates, removing SV sites with insufficient supporting samples, checking for abnormally high heterozygosity or abnormal depth patterns, correcting reference / substitution allelic orientations, and standardizing chromosome numbering and coordinate formats. For the retained sites, the minimum allele frequency, allele frequency within different varieties, deletion rate, and genotype quality distribution are calculated to form an SV site annotation table.

[0051] The SNP / InDel multi-sample VCF obtained in step S1 is integrated with the SV multi-sample VCF obtained in this step. Before integration, it is confirmed that all three types of variants use Sscrofa11.1 trunk coordinates and a uniform sample order; then, the samples are sorted by chromosome and physical location, and compressed VCF or BCF files split by chromosome are generated using bcftools concat, bcftools sort, bcftools norm, and bcftools index, or software with equivalent functions. In the integrated file, each sample has SNP, InDel, and SV genotypes simultaneously within the same reference frame.

[0052] Phased analysis was performed on the integrated multi-sample variant matrix to obtain a haplotype reference file that could be used for imputation analysis. High-density linkage disequilibrium information provided by SNPs and InDels was preserved during phasing, and SVs were used as target variant sites within the same haplotype framework. After phasing, the data was indexed and segmented by chromosome to ensure that when low-depth sequencing data or microarray data were subsequently input for the target individuals, the SV genotype could be inferred within a local window using the haplotype association relationship between SNPs / InDels and SVs.

[0053] The PGIP reference panel constructed using the above process contains 6,449 pig individuals, covering 154 breeds or populations, and integrates approximately 50.3 million SNPs, 5.8 million InDels, and 253,286 SVs. These SVs cover coding regions, intronic regions, and intergenic regions, including deletions, inversions, and copy number variations that can be represented by the reference genome, as well as insertions and presence / deletion variations that require non-reference sequences for accurate representation. Therefore, the construction of the PGIP reference panel is not a simple merging of multiple VCF files, but rather involves first establishing a unified definition of SVs in the graphical pangenome, then performing consistent SV typing on all panel individuals using the same template, and finally integrating them with SNP / InDel haplotypes to form a reference panel that can be used for imputation.

[0054] Construct a PGIP reference panel index. To facilitate subsequent filling, the PGIP is divided into multiple overlapping windows based on chromosomes. One optional implementation is to use GLIMPSE2_chunk_static or equivalent software to divide each chromosome into a 5 Mb main window, and set a 1 Mb buffer on each side of the window to retain linkage disequilibrium information near the window boundary; then use GLIMPSE2_split_reference_static or an equivalent module to cut the whole chromosome reference panel into multiple small fragments, outputting reference files and index files corresponding to each window. Each index fragment records the window coordinates, buffer coordinates, the number of SNP / InDel / SV loci included, and sample haplotype information.

[0055] For SNP microarray input data, an additional mapping table between microarray sites and PGIP sites is created before the reference panel index. Specifically, the microarray sites are converted to Sscrofa11.1 coordinates, matched with the same chromosomes and locations in the PGIP, the direction of the reference / alternative alleles is checked, A / T or C / G sites with unclear strand directions that cannot be corrected are removed, and SNP sites with clear corresponding locations are retained as microarray filling anchors.

[0056] Example 2

[0057] This embodiment assesses the saturation and population genetic characteristics of a PGIP reference panel. The specific method includes the following steps:

[0058] (1) Panel saturation assessment. Following the saturation assessment approach of the goat reference panel, multiple sample size gradients were set from the 6449 samples of PGIP, such as stratified random sampling with small, medium, and near-total sample sizes; each sample size gradient was repeated multiple times. After each sampling, the number of SNPs, InDels, and SVs that could be captured were counted, and their capture ratio relative to the entire PGIP variant set was calculated. As the sample size increased, if the number of newly added SNPs, InDels, and SVs gradually leveled off, it indicated that the reference panel had reached or was close to saturation in this type of variant.

[0059] (2) Allele frequency and deletion rate statistics. For each SV in PGIP, the minimum allele frequency, allele frequency within different breeds or populations, deletion rate, and heterozygosity rate were calculated; the number and proportion of SVs were statistically analyzed according to MAF intervals. SV loci with excessively high deletion rates, insufficient population-supporting evidence, or abnormal genotype quality were marked for use in setting quality filtering thresholds during subsequent imputation. This step demonstrates that PGIP not only includes common SVs but also retains population-specific low-to-medium frequency SVs from multiple pig breeds.

[0060] (3) SV length and type distribution analysis. SV length distributions were statistically analyzed according to insertion, deletion, inversion, duplication, copy number variation, and presence / deletion variation, and the number of SVs within different length intervals was also counted. For insertion-type SVs, insertions that can be represented by a linear reference genome and insertions dependent on non-reference paths were further distinguished; for copy number-related SVs, the distributions of different copy states or allele coding were statistically analyzed. This analysis is used to illustrate the supplementary role of the graphical pan-genome model for non-reference sequences and complex SVs.

[0061] (4) Population genetic structure analysis. Based on the filtered SV genotype matrix, loci with low deletion rates and population information were retained. PCA, neighbor-joining tree, and population structure analysis were performed using PLINK, VCFtools, Admixture, or software with similar functions. If different breeds or populations showed clustering relationships consistent with breed origin in the PCA, NJ-tree, and population structure results, it indicates that the SV genotypes in PGIP can reflect the true genetic structure of the pig population and can be used for breed identification, population differentiation analysis, and selection signal detection.

[0062] Figure 2 A shows the density distribution of different types of structural variations on the 18 chromosomes of pigs. The results show that various structural variations are widely distributed on all chromosomes. There are some differences in the density of structural variations among different chromosomes, but all have high coverage, indicating that the constructed structural variation dataset can cover the entire pig genome and effectively capture genetic variation information in the population.

[0063] Figure 2 B represents the results of the structural variation length distribution analysis. The results show that the overall length of insertion variants (INS) is greater than that of deletion variants (DEL), and both types of structural variants exhibit a wide length distribution range. Among them, most structural variants are concentrated in the 50~500 bp interval, indicating that the identified structural variants cover different length ranges and have good length representativeness and diversity.

[0064] Figure 2 C represents the statistical results of the structural variation proportions in different MAF intervals. The results show that, overall, structural variations are mainly high-frequency variations, and their proportion gradually increases with the increase of MAF. However, for deletion variations (DEL), the variation proportion in the Rare MAF interval (22.7%) is slightly higher than that in the Low MAF interval (18.5%), indicating that some deletion variations exist as rare variations in the population. Overall, different types of structural variations contain abundant low-frequency and high-frequency variation information, reflecting high genetic polymorphism.

[0065] Figure 2 D represents the principal component analysis (PCA) results based on structural variation genotypes. The results show that individuals from different populations can form clear clusters according to breed origin, effectively distinguishing Asian and European pigs in the figure. Individuals within the same population exhibit high clustering, while different populations show clear differentiation, indicating that structural variation can effectively reflect genetic differences between populations.

[0066] Figure 2 E and Figure 2 F represents the results of the neighbor-joining tree (NJ-tree) analysis based on structural variation. The results show that individuals of the same variety preferentially cluster to form independent branches, while different varieties form relatively independent genetic groups, further validating the effectiveness of structural variation data in population genetic relationship analysis.

[0067] Figure 2 G represents the results of population genetic structure analysis based on structural variation. The results show that different populations have different genetic composition characteristics, and individuals in each population can be distinguished according to their varietal origin. Furthermore, some populations share genetic components to a certain extent. These results are consistent with PCA and NJ tree analyses, further demonstrating that the obtained structural variation data can accurately characterize the genetic structure of the population.

[0068] Example 3

[0069] This embodiment describes low-depth sequencing based on PGIP and SNP microarray structural variant imputation. The specific method includes the following steps:

[0070] (1) Preparation of low-depth sequencing data for target individuals. For low-depth whole-genome sequencing input, the FASTQ data of target individuals were first processed according to the quality control method in S1 of Example 1 and aligned to the Sscrofa11.1 reference genome; then, read support information or genotype likelihood were extracted from the SNP, InDel and SV related regions covered by PGIP. The sequencing depth of target individuals can be 0.5×~2×, of which data of about 1× can be used to balance cost and fill in stability.

[0071] (2) Preparation of SNP microarray data for target individuals. For SNP microarray input, the microarray sites are first converted to Sscrofa11.1 coordinates to unify chromosome numbering and base orientation; then, they are matched with SNP sites in PGIP, removing sites where the positive and negative strand orientations cannot be determined, reference / substitute alleles are inconsistent, or sites that do not exist in PGIP. The remaining microarray sites serve as anchor points for model imputation, used to infer the SNP, InDel, and SV genotypes of the target individual in the PGIP haplotype background.

[0072] (3) Block-based imputation. Block imputation is performed on the target individual using the PGIP-phased haplotype and index files as references. Each block includes a main window and two side buffers; the main window outputs the final imputation result, and the buffers are only used to retain linkage disequilibrium information and reduce window boundary effects. For low-depth sequencing input, the imputation model utilizes the target individual's genotype likelihood, read support, and PGIP reference haplotype simultaneously; for microarray input, the imputation model mainly utilizes the haplotype linkage relationship between microarray anchor points and SNPs / InDels / SVs in PGIP.

[0073] (4) Results Merging and Quality Control. After block imputation is completed, buffer duplicate sites are removed, and the results of each chromosome window are merged into a whole genome VCF file for the target individual. INFO or equivalent imputation quality indicators are recorded for each imputation site; filtering thresholds can be set according to application needs, for example, a relatively loose threshold is used for SNPs and InDels to retain more markers, and a stricter threshold is used for SVs to improve the reliability of structural variant typing. After filtering, the number of imputation sites, deletion rate, heterozygosity rate, number of SV types, and imputation quality within different MAF intervals are statistically analyzed for each target individual.

[0074] (5) Accuracy assessment of imputation. Representative individuals with high-depth WGS or independent validation data were selected as validation samples and excluded from the PGIP reference panel during imputation. The high-depth WGS data of the validation samples were randomly downsampled to 0.5×, 1×, and 2× and used as target inputs for imputation, respectively. The high-depth WGS or independent SV detection results were used as the ground truth set, and the PGIP imputation results were used as the test set. The consistency rate and the squared Pearson correlation coefficient r were calculated. 2The concordance rate is defined as the proportion of imputed genotypes that match the true genotype, r. 2 This is used to measure the correlation between the true genotype dose and the filler dose. Accuracy can be further stratified by MAF interval, SV type, SV length, and variety origin.

[0075] (6) Application Output. The imputation results after quality control can be used for pig breed identification, genetic evaluation, population genetic structure analysis, selection signal detection, genome selection breeding, genome-wide association analysis, and quantitative trait locus analysis. Since the SV loci in PGIP have uniform coordinates and allelic codes, target individuals from different batches, different breeds, and different input data types can be jointly analyzed in the same structural variation matrix.

[0076] like Figure 3 As shown in Figure A, when performing genotype imputation based on 50K, 60K, and 80K SNP chips, the overall imputation accuracy exceeded 0.88, indicating that using medium-to-low density chips combined with a reference population can achieve relatively reliable genotype inference. With increasing sequencing depth, the imputation accuracy further improved. When the sequencing depth reached ≥1×, the accuracy exceeded 0.92, a significant improvement over chip data imputation results. This demonstrates that higher genomic information coverage helps improve the accuracy of haplotype identification and genotype inference, thus obtaining more reliable imputation results.

[0077] Figure 3 B illustrates the differences in accuracy among different varieties when using major and minor alleles for imputation. Overall, imputation accuracy based on major alleles was consistently higher than that based on minor alleles, with accuracy exceeding 0.90. This indicates that high-frequency alleles, due to their higher frequency in the population, are more easily captured by the reference haplotype, thus yielding more stable imputation results. In contrast, the accuracy of minor alleles decreased due to their lower frequency. Furthermore, significant differences existed among different varieties, with varieties D, L, and Y generally showing higher imputation accuracy than variety JH, suggesting that genetic background, linkage disequilibrium structure, and differences in the representativeness of the reference population may affect imputation results.

[0078] Figure 3C further compared the imputation accuracy of different varieties at different MAF levels. The results showed that the imputation accuracy of each variety varied with MAF, with rare variants (Rare MAF sites) showing relatively higher imputation accuracy. This may be related to the larger size of the reference population and the more complete haplotype structure in the study, allowing some rare variants to be accurately tracked and inferred. Meanwhile, a consistent trend remained among different varieties, with D, L, and Y varieties generally showing higher imputation accuracy than JH variety, further illustrating that genetic similarity between the reference and target populations is an important factor affecting imputation performance. Overall, sequencing depth, allele frequency, and variety genetic background all significantly affected genotype imputation performance, with D, L, and Y varieties showing superior imputation performance under different analytical conditions.

[0079] In summary, PGIP combines a unified definition of SVs in graphical pan-genomes, intra-panel consistency genotyping, and low-cost joint imputation analysis, enabling target individuals to obtain SV genotype information suitable for genetic analysis without high-depth whole-genome sequencing. Compared to SNP / InDel reference panels based solely on a linear reference genome, the PGIP constructed in this embodiment can supplement non-reference sequence-related SVs, reduce reference bias, and improve the consistency and generalizability of structural variation genotyping across multiple pig breeds. The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the invention. Any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention, or modify them into equivalent embodiments, without departing from the spirit and technical essence of the present invention. Therefore, any simple modifications, equivalent substitutions, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention, without departing from the content of the technical solutions of the present invention, still fall within the scope of protection of the present invention.

Claims

1. A pig structural variation reference panel based on a graphical pan-genome, characterized in that: The pig structural variation reference panel is a PGIP reference panel, containing 6,449 pig individuals from 154 breeds or populations, integrating 50.3 million single nucleotide polymorphisms (SNPs), 5.8 million insertion / deletion variants (InDel), and 253,286 structural variants (SVs). The structural variants (SVs) cover coding regions, intron regions, and intergenic regions, and can be divided into overlapping windows with buffers according to chromosomes to establish a reference panel index.

2. The pig structural variation reference panel based on graphical pangenome according to claim 1, characterized in that: The pig individuals include local breeds, commercial breeds, and their selected or hybrid populations.

3. A method for constructing a pig structural variation reference panel based on a graphical pan-genome according to claim 1 or 2, characterized in that, Includes the following steps: S1. Acquire high-depth whole-genome sequencing data of pigs from multiple breeds or populations, perform quality control, reference alignment, sorting, deduplication and variant detection on the sequencing reads, and obtain standardized sequencing data for single nucleotide polymorphisms (SNPs), insertion / deletion variants (InDel), and structural variant (SV) analysis. S2. Based on sequencing data and the pig linear reference genome Sscrofa11.1, a graphical pan-genome model containing non-reference sequences was constructed, integrating sequences missing from the linear reference genome, insertion sequences, and representative haplotype pathways from different pig individuals or populations into a unified graph structure; S3. In the graphical pangenome model, structural variations are uniformly discovered, merged, located, and allelically encoded to obtain a standardized definition system of structural variations that includes insertions, deletions, inversions, duplications, copy number variations, and presence / deletion variations; S4. Based on the standardized structural variation definition system, perform consistent structural variation genotyping on multiple pig individuals, and integrate it with SNP and InDel genotype data to construct a pig structural variation reference panel containing SNP, InDel, and SV.

4. The method for constructing a pig structural variation reference panel based on a graphical pangenome according to claim 3, characterized in that: The sequencing data described in step S1 is aligned to the Sscrofa11.1 reference genome after quality control, and SNPs and InDels are jointly detected, filtered and normalized.

5. The method for constructing a pig structural variation reference panel based on a graphical pan-genome according to claim 3, characterized in that: The graphical pan-genome model described in step S2 reduces reference bias by introducing non-reference sequences that are missing in the linear reference genome.

6. The method for constructing a pig structural variation reference panel based on a graphical pan-genome according to claim 3, characterized in that: In step S3, the structural variations are merged according to chromosome, breakpoint position, interval overlap, insertion sequence similarity, and allele path consistency, and each structural variation is assigned a unique number and a unified allele code.

7. A method for structural variation gene typing, characterized in that, Includes the following steps: (1) Obtain low-depth whole genome sequencing data or SNP chip data of the target individual and match them with the reference panel in terms of loci, coordinates and allele orientation; (2) Based on the reference panel and unified structural variation definition and coding system described in claim 1 or 2, SNP, InDel and SV joint imputation analysis is performed on the target individual to obtain the structural variation genotype results of the target individual.

8. The structural variation gene typing method according to claim 7, characterized in that: The sequencing depth of the low-depth whole genome sequencing data in step (1) is 0.5×~2×; the SNP chip data is low-density or medium-density SNP chip data; the joint imputation analysis uses haplotype linkage disequilibrium information, genotype likelihood or chip site information in the reference panel to simultaneously imput the SNPs, InDel and SV of the target individual, and filters the results according to INFO or equivalent imputation quality indicators.

9. The application of the graphical pangenome-based pig structural variation reference panel according to claim 1 or 2 in pig genetic improvement and molecular breeding.

10. The application of the structural variation gene typing method according to claim 7 or 8 in the genetic improvement and molecular breeding of pigs.

Images