Genome detection site screening method for optimizing genotype filling accuracy and application

Abstract

The application provides a genome detection site screening method for optimizing genotypic filling accuracy and an application, and belongs to the technical field of biomolecular breeding. The genotypes of multiple pig varieties are determined first, and multiple Feature values are calculated. Then, gene site division and genome filling are performed, and filling accuracy is obtained. The importance of each Feature value is calculated and rearranged by using xgboost, and finally, the site combination with the highest filling accuracy is determined. The method uses the least sites to obtain the highest efficiency, the introduction of the machine learning method xgboost significantly improves the calculation efficiency and accuracy, and lays a foundation for subsequent multiple downstream analysis.

Description

Technical Field

[0001] This invention belongs to the field of biomolecular breeding technology, specifically relating to a method and application for screening genomic detection sites to optimize the accuracy of genotype filling. Background Technology

[0002] With the rapid development of molecular breeding technology, genomic selection (GS) and genome-wide association study (GWAS) have become core methods in animal genetics and breeding. Selection strategies based on genome-wide markers can effectively improve the accuracy of trait prediction, accelerate genetic progress, and promote the modernization of breeding systems. Single nucleotide polymorphisms (SNPs) are the most commonly used genetic markers, characterized by their abundance, uniform distribution, low detection cost, and ease of standardization. They have been widely applied in genotyping, population structure analysis, QTL mapping, and genome prediction in livestock and poultry breeding.

[0003] Currently, high-density SNP microarrays (such as 50K, 60K, and 650K) offer high genotyping accuracy, but their high detection cost hinders their widespread application in large-scale populations. To reduce detection costs, breeding companies typically use low-density microarrays (such as 10K or lower) combined with genotyping imputation technology to upscale the low-density microarray data to high-density or whole-genome levels, thus balancing cost and information content. However, the selection of SNP sites in low-density microarrays significantly impacts the accuracy of subsequent imputation. Inappropriate site distribution or insufficient information will lead to decreased imputation accuracy, thereby reducing the reliability of breeding value predictions.

[0004] Traditional site selection methods often rely on single indicators such as uniform distribution of linkage disequilibrium (LD), minor allele frequency (MAF), or gene functional region characteristics. These methods fail to comprehensively consider the combined impact of multiple genomic features on imputation accuracy, resulting in limited generalization ability and predictive performance of the selected sites. Furthermore, with the accumulation of large-scale, multi-varietal genomic data, the genetic structural differences between different populations further increase the complexity of site selection optimization.

[0005] Therefore, there is an urgent need for an intelligent site selection method that can comprehensively consider multidimensional genomic feature information and aim at overall imputation accuracy. By introducing machine learning algorithms, the contribution of different feature values ​​(such as MAF, LD, gene density, conservation, functional region distribution, etc.) to imputation accuracy can be automatically evaluated, thereby achieving efficient and accurate site optimization selection, and thus obtaining the highest imputation efficiency with the fewest markers in low-density chip design.

[0006] In view of this, this invention proposes a method and application for screening genomic detection sites to optimize genotype imputation accuracy. This method utilizes genomic data from multiple pig breeds, extracts and calculates various feature values, and evaluates the contribution of each feature to imputation accuracy (SHAP value) using an xgboost model, achieving site rearrangement and screening based on feature importance. This method significantly reduces the number of required sites while maintaining high imputation accuracy, greatly improving computational efficiency and chip design quality, and providing an efficient and scalable technical foundation for subsequent genomic selection, trait prediction, and molecular breeding. Summary of the Invention

[0007] Technical Problem to be Solved: To address the aforementioned technical problems, the purpose of this invention is to provide a method and application for screening genomic detection sites to optimize genotype imputation accuracy. Multiple feature values ​​are calculated based on the genotypes of multiple pig breeds, and the overall imputation accuracy corresponding to different bins is calculated using these feature values. This screening method achieves the highest imputation accuracy with the fewest possible sites, saving sequencing and computational costs while achieving efficient and accurate imputation. The imputed whole genome can then be used for various downstream analyses such as breed identification, genome-wide association studies, genomic selection breeding, and population genetic analysis.

[0008] Technical solution: A genome detection site screening method to optimize genotype filling accuracy, which achieves efficient and accurate site selection by calculating the overall filling accuracy of different bins through Feature values.

[0009] The above-mentioned method for screening genomic detection sites to optimize genotype filling accuracy includes the following steps:

[0010] S1. Extract the genomes of multiple pig breeds and calculate the feature value and gold standard based on the genome of each breed.

[0011] S2. Divide gene loci into multiple bins (Feature_BIN) based on numerical values, and divide gene loci into multiple bins (6 or 2 Feature_BIN) based on the properties of Feature values ​​(quantitative variables or binary categorical variables).

[0012] S3. Divide the original genome into m SNP_BINs, extract one site from each SNP_BIN to form a test set of m sites, perform genome imputation and calculate the imputation accuracy;

[0013] S4. Construct an xgboost model based on Feature_BIN and the corresponding average imputation accuracy of all individuals, calculate the importance SHAP of each Feature value, and then rearrange the Feature_BIN according to the average imputation accuracy of all individuals. At this point, for each SNP, it is only necessary to multiply it with the previously calculated SHAP and rearrange it to achieve the combination of sites with the highest imputation accuracy. The formula is as follows:

[0014] (1) Input: SNP Feature → BIN Feature:

[0015]

[0016] : Represents the feature vector of the i-th single nucleotide polymorphism (SNP), which contains multiple indicators used to describe the characteristics of the SNP;

[0017] MAF: Indicates the frequency of minor alleles;

[0018] GC: Indicates GC content;

[0019] LD: Indicates the degree of chain imbalance;

[0020] (2) Output: Prediction accuracy and true fill accuracy:

[0021]

[0022]

[0023] : Represents the feature vector of the i-th SNP obtained through model f. The accuracy of the predictions obtained through forecasting;

[0024] : Indicates based on feature vectors The prediction function;

[0025] : Indicates the true accuracy of filling in the i-th SNP, that is, the actual accuracy of genotype filling;

[0026] : Represents the i-th SNP site;

[0027] (3) The contribution of the loss function to feature learning in imputation:

[0028]

[0029] L: Represents the loss function, used to measure the difference between the model's predicted value and the true value. It is expressed as mean squared error and is used to calculate the true value of all samples b. Compared with the predicted value The sum of squares between the values ​​is used to quantify the model's prediction error;

[0030] (4) SHAP value represents the contribution of Feature j to BIN b:

[0031]

[0032] GlobalScore b : Represents the global score of the b-th BIN, used to measure the overall contribution of the Feature to this BIN;

[0033] The weight of the j-th feature reflects the importance of that feature in the model;

[0034] : The SHAP value of the j-th feature to the b-th BIN, representing the quantified value of the feature's contribution to the BIN;

[0035] (5) Calculate the filling score sum for all corresponding features of each SNP, and extract a site within each SNP_BIN according to the filling score to form a site combination.

[0036] Preferably, the Feature values ​​calculated in step S1 are of eight types: Minorallele frequency (MAF), transcription start distance (DTCT), gene density (GD), gene content (GC), coding region (CDS), intron region (Intron), conserved region (Conserved), and linkage disequilibrium score (LD score, L2).

[0037] Preferably, in step S1, all sites of the genome of each variety are used for testing, and the original genotype of the high-depth genome is used as a benchmark to extract the target density sites.

[0038] Preferably, in step S3, multiple different sites are obtained by selecting sites based on Feature_BIN. The standard Beagle three-step imputation process is used to perform genome imputation based on the constructed 6000+ genome imputation panel. Then, the imputation VCF is compared with the Benchmark to calculate the imputation accuracy.

[0039] The above-mentioned method for screening genomic detection sites to optimize genotype filling accuracy is applied in animal genetic breeding.

[0040] Beneficial effects:

[0041] 1. This invention provides a method for screening genomic detection sites to optimize the accuracy of genotype filling. This method uses multiple feature values ​​to calculate the filling accuracy of corresponding site combinations, achieving the highest efficiency with the fewest sites. The filled whole genome can also meet the needs of various downstream analyses such as variety identification, genome-wide association analysis, genome selection breeding, and population genetic analysis.

[0042] 2. This invention provides a method for evaluating the importance of feature values ​​based on the machine learning algorithm xgboost. Compared with traditional evaluation methods, the goal of the xgboost algorithm is to achieve the highest global filling accuracy. By continuously learning from various feature values, the algorithm can select the feature values ​​that have the greatest impact on filling accuracy, providing a scientific basis for subsequent point selection.

[0043] 3. This invention lays the foundation for a complete genome detection process, from site design, probe design, genome detection, genome completion, to subsequent genome selection, genome-wide association analysis, variety identification, and pedigree analysis.

[0044] 4. This invention can obtain complete genomic information by accurately filling in low-depth genomes. Compared with direct high-depth sequencing, it can greatly reduce sequencing costs and meet the needs of large-scale genome breeding and selection, improving the efficiency of genetic analysis and molecular breeding of pig breeds. It is an efficient and economical solution. Attached Figure Description

[0045] Figure 1 Flowchart for the most accurate site selection method for genome sequencing;

[0046] Figure 2To assess the accuracy of filling different selection methods on the Durkin crossbred pigs in Example 3;

[0047] Figure 3 To assess the accuracy of filling different selection methods on Tunchang pigs in Example 3;

[0048] Figure 4 This is a pedigree analysis chart from Example 4 (LWH: Laiwu Black Pig; YTH: Yantai Black Pig; OTHER: Other pig breeds).

[0049] Figure 5 This is a schematic diagram of principal component analysis in Example 4;

[0050] Figure 6 The LWH pedigree content in Example 4 (red dots represent individuals with an LWH pedigree of less than 0.8);

[0051] Figure 7 The Manhattan plot of the GWAS results in Example 5 is shown. Detailed Implementation

[0052] The present invention will be further described below with reference to embodiments. These embodiments are illustrative of the present invention, but the present invention is not limited to these embodiments:

[0053] The high-density solid-phase gene chip involved in the following examples is developed based on the patented technology (CN112723304B) of Suzhou Lasso Biochip Technology Co., Ltd., and is manufactured using the core process of its microbead chip and preparation method, and is further applied to the fabrication of single nucleotide polymorphism (SNP) chips.

[0054] Unless otherwise specified, all examples were performed under standard experimental conditions, such as those described in Sambrook et al., Molecular Cloning: a Laboratory Manual (Sambrook J & Russell DW, 2001), or as recommended by the manufacturer.

[0055] Example 1

[0056] This embodiment describes sample preparation before high-density solid-phase gene chip detection, including the following steps:

[0057] I. DNA Amplification:

[0058] 1. Preparation:

[0059] (1) Set the oven to 32℃ 30 minutes in advance;

[0060] (2) Place the sample plate at room temperature to thaw, shake at 1500 speed for 1 min, and centrifuge at 2100 rpm for 1 min;

[0061] (3) Take out AMM-A, AMM-B1 and AMM-B2, let them thaw at room temperature and then immediately separate them, collecting the liquid to the bottom of the centrifuge tube; use a pipette to remove the liquid from AMM-A and add it to AMM-B1; pipette all of AMM-B2 in the tube and add it to the AMM-A centrifuge tube, rinse five times, then add the liquid to AMM-B1, invert and mix 10 times to obtain AMM; 24 samples single chip usage: take 216 μL of AMM-A, 972 μL of AMM-B1 and 27 μL of AMM-B2 into a new centrifuge tube, and invert and mix 10 times; 96 samples single chip usage: take 864 μL of AMM-A, 3888 μL of AMM-B1 and 108 μL of AMM-B2 into a new centrifuge tube, and invert and mix 10 times.

[0062] (4) Melt NES at room temperature and mix by inverting 10 times;

[0063] (5) Prepare 125 μL of 4N NaOH + 4875 μL of sterile water, and use immediately after preparation;

[0064] 2. Experimental Procedures and Records:

[0065] (1) Check if the plate number of the sample plate is the DNA used for library construction;

[0066] (2) Write the construction board number on the shallow perforated board with an oil-based pen;

[0067] (3) Take 8 μL of DNA from the sample plate into the shallow well plate (note to mark the position of A1 on the plate).

[0068] (4) Add 8 μL of 0.1N NaOH to each well, bring the pipette tip to the bottom to contact the sample, and cover with the silicone cap (note to mark the position of A1 on the membrane).

[0069] (5) Oscillate at 1500 speed for 1 min;

[0070] (6) Centrifuge at 2100 rpm for 1 min;

[0071] (7) Leave at room temperature for 10 minutes;

[0072] (8) Remove the silicone cap and add 4 μL NES; Note: After removing the silicone cap, place it face down and repeat the process in subsequent steps;

[0073] (9) Oscillate at 1500 speed for 1 min;

[0074] (10) Centrifuge at 2100 rpm for 1 min;

[0075] (11) Add 45 μL of AMM to each well;

[0076] (12) Oscillate at 1500 speed for 1 min;

[0077] (13) Centrifuge at 2100 rpm for 1 min;

[0078] (14) Place in a 32℃ hybridization box / oven for amplification, record the time, amplification time is 21 h;

[0079] (15) Register the laboratory sample tracking system and the experimental summary sheet;

[0080] II. DNA fragmentation, precipitation, resuspension, and hybridization:

[0081] 1. Preparation (DNA fragmentation):

[0082] (1) Set the heating device to 37°C;

[0083] (2) Melt DGS at room temperature and mix by inverting 10 times;

[0084] 2. Experimental Procedures and Records (DNA Fragmentation):

[0085] (1) Remove the plate and centrifuge at 280×g for 1 min;

[0086] (2) Remove the silicone cap, add 10 μL of DGS to each well, and replace the silicone cap; Note: After removing the silicone cap, place it face down and repeat the subsequent steps.

[0087] (3) Oscillate at 1500 speed for 1 min;

[0088] (4) Centrifuge at 280×g for 1 min;

[0089] (5) Incubate at 37°C for 30 min; Note: Do not turn off the heater after incubation.

[0090] 3. Preparation (Settling):

[0091] (1) FGT was left to melt at room temperature and then inverted and mixed 10 times;

[0092] (2) Prepare 100% isopropanol;

[0093] 4. Experimental Procedures and Records (Sedimentation):

[0094] (1) Remove the plate and centrifuge at 280×g for 1 min;

[0095] (2) Remove the silicone cap, add 25 μL FGT to each hole, and replace the silicone cap; Note: After removing the silicone cap, place it face up, and repeat the subsequent steps;

[0096] (3) Oscillate at 1500 speed for 1 min;

[0097] (4) Incubate at 37°C for 5 min;

[0098] (5) Centrifuge at 280×g for 1 min;

[0099] (6) Remove and discard the silicone cap, add 90 μL of 100% isopropanol to each well, replace the silicone cap, and mix the sample by inverting it at least 10 times.

[0100] (7) Place in a -20℃ refrigerator to settle for 30 min; Note: After completing this step, adjust the centrifuge to 4℃ in advance for later use;

[0101] (8) Centrifuge at 3000×g for 25 min at 4℃. After centrifugation, observe whether there is a precipitate at the bottom. The precipitate is whitish. If there is no precipitate, continue centrifugation for 10 min.

[0102] (9) Remove the centrifuge immediately after centrifugation, remove the silicone cap and discard it, and quickly pour out the supernatant; Note: If there is a delay, repeat the centrifugation at 4℃ and 3000×g for 20 min.

[0103] (10) Lay down absorbent paper in advance, invert the board and tap it a few times; Note: Do not use too much force to avoid causing sediment to fall off;

[0104] (11) Invert the plate on the blue rack and dry at room temperature for 1.5 to 2 hours until the wells are basically dry and there are no droplets. Note: If white precipitate is found to slide down during the process, dry it with the front side facing up. It is recommended to put a piece of lint-free paper under the rack to catch the falling precipitate.

[0105] 5. Preparations (Important):

[0106] (1) Turn on the sealing machine and preheat for 20 minutes;

[0107] (2) Set the hybridization chamber to 48℃;

[0108] (3) Melt RHM at room temperature, then invert and mix until no crystals remain;

[0109] 6. Experimental Procedures and Records (Re-suspended):

[0110] (1) Add 23 μL of RHM to each well (ensure that the RHM liquid covers the precipitate); Note: Avoid prolonged exposure of RHM to air, place it at -20℃ immediately after use, and do not recycle the poured-out RHM;

[0111] (2) Cover the warehouse board with aluminum foil, with the shiny side of the foil facing up. Use a sealing machine at 175°C for 2 seconds to seal the foil. Check if the aluminum foil is sealed tightly. After sealing, vibrate for 2 minutes at 1500 rpm before proceeding to the next step.

[0112] (3) Incubate in a 48℃ hybridization box / oven for 1 h; if centrifuge tubes are used, they can be resuspended on a metal bath. If the precipitate is difficult to resuspend, shake it during the resuscitation period.

[0113] 7. Preparation (hybridization):

[0114] (1) Set the hybridization chamber to 49.5℃ and 12 revolutions per minute;

[0115] (2) Set the heating instrument to 95°C 20 minutes in advance;

[0116] (3) Mix CHM by inverting it 10 times;

[0117] 8. Experimental Procedures and Records (Hybridization):

[0118] (1) Remove the construction plate and oscillate at 1500 speed for 10 min;

[0119] (2) Centrifuge at 280×g for 1 min;

[0120] (3) Place it on a 95℃ heating apparatus for denaturation for 20 min;

[0121] (4) Cool to room temperature for 15 min, then shake at 1500 rpm for 10 min. While cooling, take the high-density solid-phase gene chip out of the 4℃ refrigerator and place it at room temperature, but do not open the chip bag. Record the chip SN number.

[0122] (5) Take out the hybridization chamber, rubber pad and chip holder in advance, and put the rubber pad into the hybridization chamber and put the chip holder into the hybridization box for 10 minutes in advance;

[0123] (6) Add 400 μL CHM to the small water tank in each hybridization chamber 10 min in advance and cover it;

[0124] (7) After cooling, the well-prepared plate is centrifuged at 280×g for 1 min. After centrifugation, the plate is kept on a plate heater at 48℃ until all samples are added.

[0125] (8) Mix well by pipetting before adding samples. For 24 samples, add 13 μL of sample to the chip using a pipette. For 96 samples, add 15 μL of sample to the chip using a pipette. There should be no air bubbles on the chip after adding the sample. Do not touch the sample position when adding the sample. Add the sample carefully from the side. Observe whether the sample volume is sufficient. Place the chip in the hybridization chamber and close it. Complete the process within 1 hour.

[0126] (9) Place in a 49.5℃ (actual temperature) hybridization chamber (the temperature setting should be adjusted according to the actual situation; a maximum of 6 hybridization chambers can be placed in one hybridization chamber), record the time, hybridize for 17 hours, and obtain the product. Note: The hybridization chamber should be placed vertically, not horizontally, and should be shaken.

[0127] Example 2

[0128] This embodiment uses a high-density solid-phase gene chip to detect the sample prepared in Example 1 and obtain the target density genome. In this embodiment, 10k is used as the target density. Users can customize the target density chip design themselves, including the following steps:

[0129] 1. Preparation (chip washing and scanning):

[0130] (1) The chip constant temperature fixing device should be preheated to 44℃ 30 minutes in advance;

[0131] (2) RHM was melted at room temperature and mixed by inverting until no crystals were found; BK1 and BK2 were melted at room temperature and mixed by inverting 10 times.

[0132] (3) Melt EMM-A at room temperature, invert and mix 10 times, take 11 μL of EMM-B and add it to EMM-A, invert and mix 15 times; Amount per chip: Take 330 μL of EMM-A into a new centrifuge tube, add 2.6 μL of EMM-B, invert and mix 15 times.

[0133] (4) Melt TSM at room temperature, invert and mix 10 times, and keep away from light. Melt TCM at room temperature, invert and mix 10 times. Invert and mix DHM, CIM and WSM 10 times each.

[0134] (5) Mix SCM by inverting it 10 times. If there are small air bubbles, remove them by letting it stand, centrifuging at 280×g for 3 min or by ultrasonic degassing. It can be reused and can seal 192 chips.

[0135] (6) Dust was blown away from the assembly tank, baffle, glass tank, metal frame, glass block, chip frame, gasket, metal clip, washing tank, and plastic frame;

[0136] 2. Experimental Procedures and Records (Washing, Dyeing, Scanning):

[0137] (1) Take out the hybridization chamber prepared in Example 1 and place it at room temperature for 25-30 min;

[0138] (2) Add an appropriate amount of CIM to each of the two glass tanks;

[0139] (3) a. Carefully remove the chip; b. Remove the film from the chip, peeling it off from the lower left to the upper right, avoiding touching the sample area; c. Carefully insert the chip into the metal holder and immerse it in the CIM immersion solution, ensuring that the CIM completely covers the chip; d. Ensure that there are no foreign objects on the chip surface;

[0140] (4) Pull the metal frame up and down for 1 minute (to ensure the chip is pulled away from the liquid surface).

[0141] (5) Move the metal frame containing the chip to another glass tank containing the CIM, and pull the metal frame up and down for 1 minute (to ensure that the chip is pulled off the liquid surface).

[0142] (6) Assemble the circulation chamber as required, and temporarily immerse it in CIM after assembly;

[0143] (7) When the chip isothermal fixation device reaches 44±0.5℃, quickly install the flow chamber on the fixation device and quickly begin the following steps:

[0144] a. Add 150 μL RHM, let stand for 30 s; repeat 5 times, for a total of 6 times;

[0145] b. Add 225 μL of BK1. Let stand for 10 min; repeat once, for a total of 2 times;

[0146] c. Add 225 μL of BK2. Let stand for 10 min; repeat once, for a total of 2 times;

[0147] d. Add 300 μL EMM and let stand for 10 min;

[0148] e. Add 250 μL DHM and let stand for 1 min; repeat 2 times for a total of 3 times;

[0149] f. Let stand for 5 minutes;

[0150] g. Set the chip thermostat to 35℃ (the reagent reaction temperature is 35±0.5℃, adjust according to the temperature displayed on the thermometer), and proceed immediately;

[0151] h. Add 250 μL WSM and let stand for 1 min; repeat 5 times for a total of 6 times;

[0152] i. The chip constant temperature fixing device reaches 35±0.5℃;

[0153] j. Add 250 μL of TSM and let stand for 8 min;

[0154] k. Add 250 μL WSM and let stand for 1 min; repeat 2 times for a total of 3 times; let stand for 2 min again;

[0155] 1. Add 250 μL of TCM and let stand for 8 min;

[0156] m. Add 250 μL WSM and let stand for 1 min; repeat 2 times for a total of 3 times; let stand for 2 min again;

[0157] Add 250 μL of TSM and let stand for 8 min.

[0158] k'. Add 250 μL WSM, let stand for 1 min; repeat 2 times, for a total of 3 times; let stand for 2 min again;

[0159] Add 250 μL of TCM and let stand for 8 min.

[0160] m'. Add 250 μL WSM, let stand for 1 min; repeat 2 times, for a total of 3 times; let stand for 2 min again;

[0161] Add 250 μL of TSM and let stand for 8 min.

[0162] k''. Add 250 μL WSM, let stand for 1 min; repeat 2 times, for a total of 3 times; let stand for 2 min again;

[0163] Add 250 μL of TCM and let stand for 8 min.

[0164] Add 250 μL WSM and let stand for 1 min; repeat 2 times for a total of 3 times; let stand for 2 min again;

[0165] Add 250 μL of TSM and let stand for 8 min.

[0166] k'''. Add 250 μL WSM, let stand for 1 min; repeat 2 times, for a total of 3 times; let stand for 2 min again;

[0167] Add 250 μL of TCM and let stand for 8 min.

[0168] Add 250 μL of WSM and let stand for 1 min; repeat 2 times for a total of 3 times; let stand for 2 min again;

[0169] Add 250 μL of TSM and let stand for 8 min.

[0170] Add 250 μL of WSM and let stand for 1 min; repeat 2 times for a total of 3 times; let stand for 2 min again;

[0171] (8) Immediately remove the flow chamber from the chip constant temperature fixing device, use disassembly tools to remove the metal clips from both ends of the flow chamber, carefully remove the gasket, and avoid scratching the microbead chip area;

[0172] (9) Pour an appropriate amount of CIM into the washing tank in advance and place the chip on the plastic rack;

[0173] (10) Slowly move the plastic rack up and down 10 times (pull it away from the liquid surface) to stir the surface layer of the CIM solution and soak for 5 minutes;

[0174] (11) Pour an appropriate amount of pure water into another washing tank;

[0175] (12) Slowly move the plastic frame up and down 10 times (pull it away from the liquid surface);

[0176] (13) Pour an appropriate amount of SCM into another washing tank;

[0177] (14) Move the plastic rack containing the chip to the SCM washing tank, lift it up and down slowly twice, being careful not to create bubbles, and soak for 10 minutes;

[0178] (15) Slowly lift the plastic rack and place it on the table where the paper has been laid out, and hold it for a few seconds;

[0179] (16) Lay the chip flat on the drying rack, turn on the vacuum pump or vacuum drying oven, and dry for 30 minutes or more until the chip is completely dry;

[0180] (17) Turn on the OmniScan computer host, wait for the computer to display the desktop normally, and then turn on the OmniScan scanner.

[0181] (18) Click on the OmniScan Control Software on the computer desktop and wait for the machine to self-test and stabilize the optical path;

[0182] (19) After the software displays the initialization complete, follow the prompts to scan the chip. Note: When placing the chip, press down the four corners and confirm that the scanning parameters and configuration file are correct.

[0183] Example 3

[0184] This embodiment verifies the accuracy of site filling after xgboost site selection. High-depth genomes (10x or higher) of different breeds (Dukin crossbred pigs and Tunchang pigs) obtained in Example 2 were selected to test the site selection effect. Based on the high-depth genomes, 10k sites were extracted using the xgboost site selection method, and corresponding low-density genomes were obtained. The accuracy of filling to high depth was tested, including the following steps:

[0185] 1. The tested genomes are shown in Table 1:

[0186] Table 1. Genomic sample information for testing

[0187]

[0188] 2. Site selection method:

[0189] S1. Extract the high-depth genome of the breed of pig as a benchmark, and calculate multiple feature values ​​based on the genome;

[0190] S2. Based on the properties of Feature values ​​(quantitative variables or binary categorical variables), divide gene loci into multiple bins (6 or 2 Feature_BINs).

[0191] S3. Divide the original genome into 10k SNP_BINs, extract the site with the highest GlobalScore from each SNP_BIN to form a 10k site test set, use the standard Beagle three-step imputation process to imput the genome, and then compare the imputation VCF with the Benchmark to calculate the imputation accuracy.

[0192] S4. Based on the importance SHAP of each Feature value, and then according to the Feature_BIN corresponding to the filling accuracy, for each SNP, we only need to multiply it with the previously calculated SHAP and rearrange it to achieve the combination of sites with the highest filling accuracy. The formula is as follows:

[0193] (1) Input: SNP Feature → BIN Feature:

[0194]

[0195] : Represents the feature vector of the i-th single nucleotide polymorphism (SNP), which contains multiple indicators used to describe the characteristics of the SNP;

[0196] MAF: Indicates the frequency of minor alleles;

[0197] GC: Indicates GC content;

[0198] LD: Indicates the degree of chain imbalance;

[0199] (2) Output: Prediction accuracy and true fill accuracy:

[0200]

[0201]

[0202] : Represents the feature vector of the i-th SNP obtained through model f. The accuracy of the predictions obtained through forecasting;

[0203] : Indicates based on feature vectors The prediction function;

[0204] : Indicates the true accuracy of filling in the i-th SNP, that is, the actual accuracy of genotype filling;

[0205] : Represents the i-th SNP site;

[0206] (3) The contribution of the loss function to feature learning in imputation:

[0207]

[0208] L: Represents the loss function, used to measure the difference between the model's predicted value and the true value. It is expressed as mean squared error and is used to calculate the true value of all samples b. Compared with the predicted value The sum of squares between the values ​​is used to quantify the model's prediction error;

[0209] (4) SHAP value represents the contribution of Feature j to BIN b:

[0210]

[0211] GlobalScore b : Represents the global score of the b-th BIN, used to measure the overall contribution of the Feature to this BIN;

[0212] The weight of the j-th feature reflects the importance of that feature in the model;

[0213] : The SHAP value of the j-th feature to the b-th BIN, representing the quantified value of the feature's contribution to the BIN;

[0214] (5) Calculate the sum of the filling scores for all corresponding features of each SNP, and extract a site within each SNP_BIN according to the filling score to form a site combination;

[0215] S5. Test results were obtained using a single feature point selection method, and the filling accuracy was compared with that of the XGBoost point selection method.

[0216] like Figure 1The diagram shown is a flowchart of the testing process for the most accurate site selection method for genome detection.

[0217] like Figures 2-3 As shown, the accuracy results of different spot selection methods for imputation are displayed for two pig breeds (Duginosaur crossbred and Tunchang pig). In Dugin crossbred pigs, the accuracy of imputation based on single features (CDS, Conserved, DTCT, GC, GD, and Intron) is low, all below 0.94; the accuracy of imputation based on L2 and MAF methods shows a large difference, with L2_bin_4 and MAF_bin_3 showing higher accuracy, exceeding 0.94; the spot selection method calculated based on xgboost has the highest imputation accuracy, at 0.9449 (…). Figure 2 In Tunchang pigs, the accuracy of fill-in methods based on single features (CDS, Conserved, DTCT, GC, GD, and Intron) was low, all below 0.925; the fill-in accuracy based on L2 and MAF showed significant differences, with L2_bin_4 and MAF_bin_6 showing higher accuracy; the fill-in accuracy based on the point selection method calculated by xgboost was the highest, at 0.9396. Figure 3 ).

[0218] Example 4

[0219] This embodiment describes a pedigree analysis of Shandong Laiwu pigs based on XGBoost 10k sampling points, verifying the accuracy of breed identification by combining pedigree records and phenotypic records. The steps include:

[0220] 1. Genome sequencing: Genome sequencing data of 137 Shandong Laiwu pigs were available, and sequencing was performed using a chip designed with 10k selected sites based on XGBoost.

[0221] 2. Pedigree analysis: First, the genome was analyzed based on the R package GBC 1.0 for 124 pig breeds in the iDIGS database. After screening the top-ranked pig breeds based on the results, Laiwu Black Pig (LWH), Yantai Black Pig (YTH) and some other pig breeds were used as the pedigree sources for further pedigree analysis.

[0222] 3. Principal Component Analysis (PCA): Plot the first and second principal components.

[0223] 4. Screening for purebred / abnormal individuals.

[0224] like Figure 4 As shown, the results indicate that the Laiwu Black Pig lineage is dominant, which is in line with expectations.

[0225] like Figure 5As shown, PCA principal component analysis revealed that the clustering of these 137 individuals was biased towards LWH (Laiwu Black Pig), proving that the bloodlines of most pigs were not mixed, which is in line with expectations; however, some individuals deviated significantly, suggesting mixed bloodlines, which is consistent with breed records.

[0226] like Figure 6 As shown, based on past experience, individuals with an LWH (Laiwu Black Pig) pedigree content above 80% are considered highly reliable; those above 70% but below 80% are considered reliable; and those below 70% are considered suspected. Analysis of LWH (Laiwu Black Pig) pedigree content revealed 13 individuals with an LWH (Laiwu Black Pig) pedigree content below 70%, suggesting possible mixed pedigrees. Based on these results, individuals with an LWH (Laiwu Black Pig) pedigree content above 70% (reliable) and 80% (highly reliable) are selected for future breeding plans.

[0227] The above results are consistent with individual records and pedigree information, demonstrating that the chip designed using the XGBoost point selection method can meet the needs of variety identification.

[0228] Example 5

[0229] This embodiment performs GWAS analysis based on the imputed genome, including the following steps:

[0230] 1. Study Subjects: A total of 3724 pigs (see Table 2) were selected, born between 2017 and 2023, and raised in farms in Anhui Province (N=1014) and Guangxi Zhuang Autonomous Region (N=2710). The interquartile range (IQR) was used to remove outliers, eliminating samples with values ​​less than Q1~1.5×IQR and greater than Q3+1.5×IQR, ultimately resulting in 3635 pigs: 1358 Duroc, 647 Landrace, 1132 Yorkshire, and 498 Pietrain.

[0231] Table 2. Sample collection status of five imported pig breeds from Shiji Biotechnology Co., Ltd.

[0232]

[0233] Live pigs were scanned with CT scans, and four phenotypic data were obtained after image processing and calculation: lean meat percentage (LMP), fat percentage (FP), bone percentage (BP), and number of ribs (NR).

[0234] 2. Genome imputation: After extracting 10,000 genomic loci using the xgboost site selection method, the original genome detection results were imputed using genome imputation software (Beagle) and a high-depth genome imputation panel to obtain high-density genome data;

[0235] 3. GWAS Analysis: First, the heritability of each trait was estimated using a single-trait model, and then the genetic associations between traits were estimated using a two-trait model. The GWAS model was performed using a linear mixture model in GEMMA software, with Bonferroni correction, and the significance threshold was set to P < 2.42 × 10⁻⁶. −7 The results are presented in the form of a Manhattan plot.

[0236] like Figure 7 As shown, from top to bottom, the GWAS results for lean meat percentage (LMP), bone percentage (BP), and number of ribs (NR) are displayed. This shows that the strategy of extracting sites based on the xgboost site selection method and then filling in the gaps can meet the needs of our scientific research on the analysis of genetic mechanisms. GWAS can obtain significant sites and locate key genes of traits.

[0237] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. 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, shall still fall within the scope of protection of the present invention.

Claims

1. A method for screening genomic detection sites to optimize genotype filling accuracy, characterized in that, This method achieves efficient and accurate site selection by calculating the filling accuracy of corresponding sites using Feature values. The specific method includes the following steps: S1. Extract the genomes of multiple pig breeds and calculate the Feature value and the gold standard Benchmark based on the genome of each breed; S2. Based on numerical values, gene loci are divided into multiple Feature_BINs, and based on the properties of Feature values, gene loci are divided into 6 or 2 Feature_BINs; S3. Divide the original genome into m SNP_BINs, extract one site from each SNP_BIN to form a test set of m sites, perform genome imputation and calculate the imputation accuracy; S4. Construct an xgboost model based on Feature_BIN and the corresponding average imputation accuracy of all individuals, calculate the importance SHAP of each Feature value, and then rearrange Feature_BIN according to the average imputation accuracy of all individuals. At this point, for each SNP, it is only necessary to multiply it with the previously calculated SHAP and rearrange it to achieve the combination of sites with the highest imputation accuracy. The formula is as follows: (1) Input: SNP Feature → BIN Feature: ； : Represents the feature vector of the i-th single nucleotide polymorphism (SNP), which contains multiple indicators used to describe the characteristics of the SNP; MAF: Indicates the frequency of minor alleles; GC: Indicates GC content; LD: Indicates the degree of chain imbalance; (2) Output: Prediction accuracy and true fill accuracy: ； ； : Represents the feature vector of the i-th SNP obtained through model f. The accuracy of the predictions obtained through forecasting; : Indicates based on feature vectors The prediction function; : Indicates the true accuracy of filling in the i-th SNP, that is, the actual accuracy of genotype filling; : Represents the i-th SNP site; (3) The contribution of the loss function to feature infilling: ； L: Represents the loss function, used to measure the difference between the model's predicted value and the true value. It is expressed as mean squared error and is used to calculate the true value of all samples b. Compared with the predicted value The sum of squares between the values ​​is used to quantify the model's prediction error; (4) SHAP value represents the contribution of Feature j to BIN b: ； GlobalScore b : Represents the global score of the b-th BIN, used to measure the overall contribution of the Feature to this BIN; The weight of the j-th feature reflects the importance of that feature in the model; : The SHAP value of the j-th feature to the b-th BIN, representing the quantified value of the feature's contribution to the BIN; (5) Calculate the filling score sum for all corresponding features of each SNP, and extract a site within each SNP_BIN according to the filling score to form a site combination.

2. The method for screening genomic detection sites to optimize genotype filling accuracy according to claim 1, characterized in that: The Feature values ​​calculated in step S1 are of eight types: minor allele frequency, transcription start site distance, gene density, gene content, coding region, intron, conserved region, and linkage disequilibrium fraction.

3. The method for screening genomic detection sites to optimize genotype filling accuracy according to claim 1, characterized in that: In step S1, all sites of the genome of each variety are tested, and the original genotype of the high-depth genome is used as a benchmark to extract the target density sites.

4. The method for screening genomic detection sites to optimize genotype filling accuracy according to claim 1, characterized in that: In step S3, multiple different sites are obtained by selecting sites based on Feature_BIN. The standard Beagle three-step imputation process is used to imput the genome based on the constructed 6000+ genome imputation panel. Then, the imputation VCF is compared with the Benchmark to calculate the imputation accuracy.

5. The application of the genome detection site screening method for optimizing genotype filling accuracy as described in any one of claims 1 to 4 in animal genetic breeding.

Images