Method for constructing a set of SNP sites for identifying sasheng pigs and application thereof

Abstract

The present application relates to a kind of methods for constructing identification SNP site set of Sazba pig and its application, comprising the following steps: S1, the SNP site breed score information of Sazba pig is calculated;S2, Sazba pig SNP site is selected from above-mentioned SNP site, using K-means clustering method is divided into three cluster clusters;S3, the SNP site of three cluster clusters is further selected to obtain Sazba pig identification SNP site set candidate site ranking;S4, the first 30-40 sites of candidate site selection are used as the identification SNP site set of Sazba pig.The method for constructing identification SNP site set of Sazba pig and its application provided by the present application can quickly identify Sazba pig from multiple pig strains, or quickly identify Sazba pork from multiple meat products to be tested, and the theoretical accuracy of identification is as high as 99.21%.The present application has obvious Sazba pig species specificity, stronger practicability, and good development and application prospect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of animal breed identification, and particularly relates to a method for constructing a SNP site set for identifying Sa Bapig and application thereof. BACKGROUND

[0002] Sa Bapig is a local pig breed in Yunnan Province, mainly produced in Luqian Yi and Miao Autonomous County under Kunming and Wuding, Chuxiong Yi Autonomous Prefecture, Duanzhong, Luofeng, Yaoan, Dayao, Shuangbai, etc., and named after Sa Yingchang Town (formerly known as Sa Bashi) in Luqian County. It has the advantages of high reproductive capacity, tolerance to coarse feed, good disease resistance, and good meat quality, etc. It is a good local breed in Yunnan Province and is evaluated as one of the "six famous pigs" in Yunnan Province by the Department of Agriculture of Yunnan Province. Sa Bapig has a delicious taste and is sold at a higher price than ordinary pig breeds. There are often cases of other pig breeds pretending to be Sa Bapig. In order to maintain the unique and pure bloodline of Sa Bapig and avoid other pig populations or pork products replacing Sa Bapig in the market, it is necessary to quickly and accurately identify the pig breeds or pork to be tested. Therefore, how to achieve the rapid and accurate identification of Sa Bapig is a direction that needs to be developed at present, and developing a method for quickly and reliably identifying Sa Bapig is conducive to the protection, exploration and breeding of Sa Bapig genetic resources.

[0003] Gene Identification Card mainly uses DNA fingerprinting technology to select a number of fixed sites for identification. Genes are inherent and unchanging genetic markers inherited from parents by offspring. Only a few SNP sites can be identified. The combination of these selected SNP sites has uniqueness. Chinese patent 201810310363.3 discloses Guanzhuang Meishan pig SNP sites, SNP chip, detection primer combination, detection kit, application and germplasm identification method; Chinese patent 202110782383.2 discloses a pig family division and parentage identification kit and method, which shows that SNP sites have very strong specificity and are crucial for breed identification. However, the existing technology often uses PCR amplification followed by sequencing for identification, which is time-consuming and costly. Chinese patent 202211692300.1 discloses a Jinhua pig identification method, which collects specific genetic label combinations of Jinhua pigs, performs whole genome sequencing, and performs whole genome association analysis by linear mixed model to screen the top 25 SNP sites with significant values to form an identification data set; and uses the identification data set to construct a machine learning model for identification. This identification method uses a mathematical model, which is simple and fast to operate, but since the SNP site data sets for identifying different pig breeds are different, this method cannot be used to identify Sa Bapig. Constructing a gene identification card for local pig breeds can effectively protect and develop local pig breeds, but so far, there is no report on SNP site data sets for Sa Bapig breed identification, their construction methods and applications.

[0004] Zhao C, Wang D, Teng J et al. published an article in the journal Journal of Animal Science and Biotechnology, 2023, 14(1): 85, Breed identification using breed-informative SNPs and machine learning based on whole genome sequence data and SNP chip data [J], discloses a breed information detection and breed allocation method for cattle, which is to calculate breed information score, score sorting, select the site with higher score, and then perform machine learning breed prediction, but the result is good or bad depends on the size of the original training set population, and a large number of site numbers need to be selected to improve the accuracy to more than 99%.

[0005] At present, there is no report on the construction method and application of accurate identification SNP site set for Saba pigs. SUMMARY

[0006] In order to solve the above problems, the present application provides a method for constructing a Saba pig identification SNP site set and its application.

[0007] The present application provides a method for constructing a Saba pig identification SNP site set, which comprises the following steps:

[0008] S1, calculating the SNP site breed score information of Saba pigs: collecting Saba pig blood samples for whole genome sequencing, and summarizing various pig breed SNP data with IAnimal database, calculating the breed score of all pig breed SNP sites, that is, the genetic difference Delta, the genetic differentiation index Fst between paired populations and the allocation information degree In;

[0009] S2, selecting Saba pig SNP sites with unique allele genotype and Delta>0.49252 or Fst>0.14251 or In>0.77959 from the above SNP sites, and standardizing the score value of each SNP site, and classifying according to the standardized score value of each SNP site using K-means clustering method, which is divided into three clusters, wherein the Fst score of the sites in cluster 1 is higher than the average, and the Delta and In scores are slightly higher than the average; the Fst score of the sites in cluster 2 is higher than the average, and the Delta and In scores are lower than the average; the Fst score of the sites in cluster 3 is lower than the average, and the Delta and In are higher than the average;

[0010] S3, for the SNP sites of the three clusters obtained in S2, the following algorithm is adopted for further selection:

[0011] (1) Calculate the central site common to SNP sites in three clustering clusters

[0012]

[0013] wherein, is the central site, X ij represents the coordinate of the i-th site in the j-th dimension, n is the total number of sites,

[0014] j = 3;

[0015] (2) Calculate the distance D of each SNP site in each clustering cluster from the central site:

[0016]

[0017] (3) Iterative selection of sites:

[0018] According to the order of D from large to small, SNP sites are selected from cluster 1, cluster 2 and cluster 3 in turn, and are sequentially and without replacement included in the site selection area. Three sites are selected from cluster 1, cluster 2 and cluster 3 in turn in each round, and the process is repeated until all sites are selected, thereby reordering the original sites. The order is the candidate site ordering of the Saziba pig identification SNP site set.

[0019] S4, reselection of candidate sites: construct a LightGBM model, select the top 30-40 sites from the above Saziba pig identification SNP site set candidate sites as the Saziba pig identification SNP site set.

[0020] Preferably, in S4, the top 30 sites are selected from the above Saziba pig identification SNP site set candidate sites as the Saziba pig identification SNP site set.

[0021] Further, the K-means clustering uses the R package NbClust, and the method is as follows: sc.km <- kmeans(sc_nl, 3, nstart = 25), wherein sc.km is the final classification list; the kmeans() function is the K-means clustering function of the NbClust package; sc_nl is the scoring table of the SNP breed score; the parameter 3 indicates that it is classified into three categories; the nstart = 25 parameter indicates that 25 centroid presets are generated to ensure stable classification results.

[0022] The application further provides a SNP site set for identifying Sutai pigs, which is constructed by the method and contains 30 SNP sites, namely: SNP number 1116156, genotype TT, located at 195559511 of chromosome 1; SNP number 897074, genotype CC, located at 148210111 of chromosome 1; SNP number 4079088, genotype AA, located at 75294722 of chromosome 4; SNP number 950547, genotype AA, located at 160129325 of chromosome 1; SNP number 897070, genotype GG, located at 148210051 of chromosome 1; SNP number 4079094, genotype CC, located at 75294931 of chromosome 4; SNP number 862065, genotype AA, located at 141987098 of chromosome 1; SNP number 980377, genotype TT, located at 165070451 of chromosome 1; SNP number 4082678, genotype GG, located at 75639414 of chromosome 4; SNP number 950601, genotype TT, located at 160136482 of chromosome 1; SNP number 897072, genotype CC, located at 148210085 of chromosome 1; SNP number 4082685, genotype CC, located at 75640471 of chromosome 4; SNP number 953025, genotype TT, located at 160600275 of chromosome 1; SNP number 980650, genotype CC, located at 165118874 of chromosome 1; SNP number 4083231, genotype CC, located at 75755375 of chromosome 4; SNP number 953861, genotype CC, located at 160708948 of chromosome 1; SNP number 977888, genotype GG, located at 164637654 of chromosome 1; SNP number 4079656, genotype AA, located at 75343352 of chromosome 4; SNP number 11929687, genotype GG, located at 116110241 of chromosome 13; SNP number 981803, genotype GG, located at 165438294 of chromosome 1; SNP number 4082677, genotype TT, located at 75639169 of chromosome 4; SNP number 953388, genotype TT, located at 160641713 of chromosome 1; SNP number 935626, genotype GG, located at 156644321 of chromosome 1; SNP number 4082666, genotype GG, located at 75635325 of chromosome 4;SNP number 938563, genotype AA, located at 1 chromosome 157479708; SNP number 981840, genotype AA, located at 1 chromosome 165445915; SNP number 4079084, genotype AA, located at 4 chromosome 75294571; SNP number 727571, genotype TT, located at 1 chromosome 109395215; SNP number 896836, genotype AA, located at 1 chromosome 148184989; SNP number 4083278, genotype GG, located at 4 chromosome 75759399.

[0023] The application provides a kind of application of identifying SNP site set of Saba pig in identifying Saba pig breed, the application method is, collecting blood sample of pig species to be measured for whole genome sequencing, according to the position information of the above-identified SNP site set, obtain the genotype information of each pig species to be measured corresponding site, constitute test sample set, compare the genotype of 30 sites of the above-identified SNP site set with test sample set, if genotype is completely identical, it can be judged that the pig species to be measured is Saba pig, otherwise it is other breed.

[0024] The beneficial effects of the application are as follows:

[0025] The application provides a kind of method for constructing Saba pig identification SNP site set and its application, which can quickly identify Saba pig from multiple pig breeds, or quickly identify Saba pork from multiple meat products to be measured. The construction method of the application selects 30 sites in order as the identification SNP site set of Saba pig, and the theoretical accuracy of identification is as high as 99.21%. The application has obvious Saba pig breed specificity, strong practicability, and good development and application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 It is the clustering effect diagram of example 1;

[0027] Figure 2 It is the standard score feature comparison diagram of three clusters of example 1;

[0028] Figure 3 It is the site number-accuracy fold line graph of example 1;

[0029] Figure 4 It is the distribution and density diagram of the first 30 sites on pig chromosome of example 1;

[0030] Figure 5 It is the site number-accuracy fold line graph of comparative example. DETAILED DESCRIPTION

[0031] The application will be further described in conjunction with the following examples.

[0032] Example 1: Constructing the SNP site set for identifying Sa Bapigs

[0033] S1, calculate the SNP breed score information of all pig populations:

[0034] S1.1, population composition: 100 Sa Bapigs were sampled from Yuanjia Village, Ziwu Town, Chuxiong City, Yunnan Province, 80 Dahu pigs were sampled from Fuyuan County, Yunnan Province, and 45 Diqing Tibetan pigs were sampled from Shixiang Village, Jinjiang Town, Shangri-La City, Yunnan Province. Other 45 kinds of pigs are 69 Tibetan pigs, 48 Meishan pigs, 23 Erhuamaian pigs, 20 Anqing Liubai pigs, 20 Wannan black pigs, 17 Min pigs, 14 Korean black pigs, 12 Laiwu pigs, 11 Rongchang pigs, 7 Wuzishan pigs, 6 Hetao large-ear pigs, 6 Lulei black pigs, 6 Tongcheng pigs, 6 Bama Xiang pigs, 6 Luchuan pigs, 5 Jiangquhai pigs, 4 Jinhua pigs, 4 Xiang pigs, 3 Wujin pigs, 3 Neijiang pigs, 3 Enshi black pigs, 3 Penzhou mountainous pigs, 2 Wannan spotted pigs, 2 Songliao black pigs, 2 Leping pigs, 2 Dian'nansmall-ear pigs, 188 Duroc pigs, 122 Yorkshire pigs, 56 Landrace pigs, 30 Pietrain pigs, 20 Miniature small Xiang pigs, 14 Yucatan miniature pigs, 14 Berkshire pigs, 8 curly wild boars, 8 Iberian pigs, 5 Mangalitza pigs, 5 Hampshire pigs, 3 Tamworth pigs, 2 Middle White pigs, 2 Gloucestershire Old Spots pigs, 2 Creole pigs, 2 Spanish Murciana pigs, 2 Castellana pigs, 2 Wessex Saddleback pigs, and 2 Angler saddleback pigs. The above Sa Bapigs, Dahu pigs, and Diqing Tibetan pigs were self-blooded for whole genome sequencing, and the SNP data of the other 45 kinds of pigs came from the IAnimal database https: / / ianimal.pro / reception / genome / population;

[0035] S1.2, data processing:

[0036] (1) Whole genome resequencing: about 0.3-0.5 mL of animal blood samples were collected from each individual of the above Sa Bapigs, Dahu pigs, and Diqing Tibetan pigs, and stored at -40℃. DNA samples were extracted using EasyPure Blood Genomic DNA Kit and other methods. The quality and concentration of the extracted DNA were detected by Nanodrop nucleic acid protein analyzer. The A260 / 280 ratio was 1.8-2.0, and the concentration was ≥50 ng / μL to determine the pass. The samples with qualified concentration were subjected to whole genome sequencing on the DNBSEQ platform.

[0037] (2) Genotyping and filling: Using Sscrofa11.1 as the reference genome, the above sequencing results were genotyped using the GATK standard process, and BCFtools and IAnimal database were used to merge each pig species site. In order to obtain high confidence difference sites, VCFtools was also used for filtering, shapeit5 was used for phasing, and impute5 was used for genotype filling. Finally, the data of 1016 samples containing 16,167,605 sites were obtained;

[0038] (3) Calculate breed score of all sites: Use TRES to calculate the genetic difference Delta, the genetic differentiation index Fst between paired populations, and the distribution information score In of all sites;

[0039] S2, preliminary selection of sites and clustering:

[0040] S2.1, acquisition of key sites: In order to shorten the data processing time, among all the above sites, we selected sites with high breed score, site-specific score threshold Delta>0.49252, Fst>0.14251 or In>0.77959, and part of the sites with unique allele genotype in Sazhe pigs, and standardized processing according to the score value of each site.

[0041] S2.2 Clustering: K-means clustering method was used to classify high-score sites, and a total of three types of sites were classified. K-means clustering used the R package NbClust, and the method was as follows: sc.km<-kmeans(sc_nl,3,nstart=25). Among them, sc.km is the final classification list; the kmeans() function is the K-means clustering function of the NbClust package; sc_nl is the scoring table of the above SNP breed score; the parameter 3 indicates that it is classified into three categories; the nstart=25 parameter indicates that 25 centroid preset values are generated to ensure that the clustering result is stable. The clustering effect is shown in Figure 1 .

[0042] The K-means clustering algorithm divides the above selected sites into different three categories, as shown in Figure 2As shown, the standard score characteristics of the three clusters can be explained as follows: ① in the loci of cluster 1, the Fst score is higher than the average of all selected loci, and the Delta and In scores are slightly higher than the average; ② in the loci of cluster 2, the Fst score is higher than the average, and the Delta and In scores are lower than the average; ③ in the loci of cluster 3, the Fst score is lower than the average, and the Delta and In scores are higher than the average. Assuming that each cluster of loci contributes equally to the identification of Saba pigs, it is necessary to equally select loci from the three clusters, that is, three loci are selected from cluster 1, cluster 2 and cluster 3 in turn in each round, and the process is repeated until a sufficient set of loci for identifying Saba pigs is obtained.

[0043] S2.3 For the above three types of loci, the following algorithm is adopted to select the loci:

[0044] (1) Calculate the central loci of the SNP loci in the three clusters

[0045]

[0046] wherein, is the central loci, X ij represents the coordinates of the ith locus in the jth dimension, and n is the total number of loci,

[0047] j = 3;

[0048] (2) Calculate the distance D of each SNP locus in each cluster from the central locus:

[0049]

[0050] (3) Iterative selection of loci:

[0051] According to the order from large to small, SNP loci are selected from cluster 1, cluster 2 and cluster 3 in turn, and are sequentially and without replacement included in the locus selection area. In each round, three loci are selected from cluster 1, cluster 2 and cluster 3 in turn and the process is repeated, until all loci are selected, thereby reordering the original loci. The order is the candidate locus order of the Saba pig identification SNP locus set. In this embodiment, a total of 104 SNP loci are obtained:

[0052] Table 1 Order of candidate loci of Saba pig identification SNP locus set obtained in Example 1

[0053]

[0054]

[0055]

[0056] S4, reselection of candidate sites: build a LightGBM model, for the above candidate sites, sequentially select 3, 4, 5, 10, 15, 20, 25, 30, 35 and 40 sites, and through ten-fold cross-validation method, make variety prediction, compare the output results of the model with the reference results, and calculate the accuracy. The model building method is as follows:

[0057] LightGBM model training and evaluation are carried out through autogluon package in python, and the training model is:

[0058] predictor=

[0059] TabularPredictor(label=label).fit(train_data=df_train, hyperparameters={'GBM':{}}).

[0060] Among them, TabularPredictor is the model trainer in autogluon package, label is the breed name of training set, fit method indicates specific parameter setting: train_data=df_train represents the above Saba pig SNP site candidate set, hyperparameters={'GBM':{}} indicates that LightGBM is set as the training model of machine learning. The accuracy calculation method is as follows:

[0061]

[0062] Among them, A is the accuracy, TP represents true positive, that is, the number of samples correctly predicted as Saba pig;

[0063] TN represents true negative, that is, the number of samples correctly predicted as other varieties; FP represents false positive, that is, the number of samples incorrectly predicted as Saba pig; FN represents false negative, that is, the number of samples incorrectly predicted as other varieties.

[0064] Table 2 comparison of prediction accuracy of different site numbers

[0065] Number of sites True positives True negatives False positives False negatives Accuracy 3 100 888 28 0 97.24％ 4 100 888 28 0 97.24％ 5 100 888 28 0 97.24％ 10 100 896 20 0 98.03％ 15 100 896 20 0 98.03％ 20 100 902 14 0 98.62％ 25 100 905 11 0 98.92％ 30 100 908 8 0 99.21％ 35 100 908 8 0 99.21％ 40 100 908 8 0 99.21％

[0066] As shown in Tables 2 and Figure 3 Under the condition of as high accuracy as possible and as few sites as possible, 30-40 sites are selected as Saba pig identification SNP site set according to the above method, and the theoretical accuracy of identification is as high as 99.21%, and the theoretical accuracy of 30 sites, 35 sites and 40 sites is 99.21%.

[0067] Based on the above conclusions, the first 30 SNP loci for identification of Saba pigs were selected, as shown in Table 3:

[0068] Table 3 shows the set of SNP loci selected from the top 30 loci for identification in Saba pigs.

[0069]

[0070]

[0071] from Figure 4 The distribution and density map of the first 30 loci on pig chromosomes shows that the above 30 loci are mainly distributed on pig chromosomes 1, 4 and 13, with the majority distributed on chromosome 1.

[0072] Example 2: Application of the SNP locus set for Saba pig identification in practical identification

[0073] SNP data from another 10 Saba pig samples and 92 samples from other pig breeds were collected and combined. Based on the SNP locus set of the aforementioned 30 Saba pig identification sites, genotype information for the corresponding loci of each pig was obtained from the whole genome data. The most frequently occurring major alleles were selected to form the test sample set. By comparing the genotypes of the test sample set with the genotypes of the aforementioned 30 Saba pig identification SNP locus set, if the genotypes were completely identical, a sample could be identified as a Saba pig; otherwise, it was another breed. The comparison results are shown in Table 4.

[0074] The specific implementation steps are as follows:

[0075] The samples can be of any pig breed, but SNP information for that sample must be obtained. In this embodiment, the SNP information for the Saba pig, Dahe pig, and Diqing Tibetan pig samples in the test sample set was obtained from whole-genome sequencing of blood samples collected independently. Specifically, approximately 0.3-0.5 mL of blood sample was collected from each individual of Saba pig, Dahe pig, and Diqing Tibetan pig, frozen at -40℃, and nucleic acids were extracted from the DNA samples using methods such as the EasyPure Blood Genomic DNA Kit. The extracted DNA was then subjected to quality testing and concentration determination using a Nanodrop nucleic acid and protein analyzer. An A260 / 280 ratio of 1.8-2.0 and a concentration ≥50 ng / μL were considered acceptable. Samples meeting the concentration standard were then subjected to whole-genome sequencing on the DNBSEQ platform.

[0076] The SNP information of the test sample set of other pig breeds was from the IAnimal database, which was combined with the sequencing data of Saba pig, Dikei pig and Dikei Tibetan pig and genotyped to form a test sample set. Specifically, the sequencing results of Saba pig, Dikei pig and Dikei Tibetan pig were combined with the SNP data information of each pig breed in the IAnimal database, and the reference genome was Sscrofa11.1. The genotyping was performed using the GATK standard process, the SNP sites were combined using bcftools, and the phase was determined using shapeit5, and the filling was performed using impute5.

[0077] The genotyping results of each sample in the test sample set were compared with the genotypes of the 30-site Saba pig identification SNP site set in Example 1. If the genotypes of the former and the latter are completely consistent, it indicates that the sample is derived from Saba pig, otherwise it is other individual. Specifically:

[0078] (1) The genotype information of 30 sites of all test samples was extracted using bcftools: bcftools filter -R 30_sites.txt > target.vcf; wherein, filter is the filtering function of bcftools; vcf.gz is the compressed vcf file after combining; -R is the parameter for reading the file; 30_sites.txt is the information of 30 sites, the first column is the chromosome number, the second column is the physical distance, separated by tab; target.vcf is the specific genotype information of 30 sites;

[0079] (2) Using plink, convert target.vcf to PED format: plink --vcf target.vcf --recode --out target; wherein, --vcf is the parameter for reading the vcf file; --recode --out is the output parameter; target is the file name prefix of the output file, and the output file includes target.ped, target_map;

[0080] (3) Using awk, extract the genotype information of each sample: awk 'NR==FNR{printf $1 "_" $4 "\t"} NR>FNR{if(FNR==1){print""}; for(i=1;i<=30;i++){printf "%s%s\t",$(i*2+5),$(i*2+6)}; printf "%s\n", $1}' target_map target_ped > target_30.txt; target_30.txt is the genotype file of each sample, each column except the last column is the genotype of the site, and the last column is the sample breed;

[0081] (4) Compare the information of target_30.txt with the identified SNP site set of 30 sites of Saba pigs, use excel to summarize, and obtain Table 4, i.e., the constructed verification population and verification results:

[0082] Table 4: Results of whether the to-be-tested pigs are Saba pigs according to the identified SNP site set of Saba pigs

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092]

[0093] As shown in Table 4, in this embodiment, the identified SNP site set of Saba pigs is used, and in 31 pig breeds, a total of 102 to-be-tested samples, Saba pigs can be accurately identified at 100%.

[0094] Comparative Example: Comparison with prior art

[0095] Zhao C, Wang D, Teng J et al. published an article in the journal Journal of Animal Science and Biotechnology, 2023, 14(1): 85, entitled "Breed identification using breed-informative SNPs and machine learning based on whole genome sequence data and SNP chip data [J]", which discloses a method for detecting breed information and breed allocation of cattle. The method is similar to the method of the present application, but when constructing the SNP site set for identification, the selection method of the site is to calculate the breed score of all individuals of the breed, sort from large to small, and screen high-score sites. It does not screen for single-site genotypes. Because the genotype is not determined, the genotype information of the sample to be tested needs to be input into the calculated model for prediction to verify the result. The result depends on the size of the original training set. After the population changes, subsequent training needs to be supplemented to ensure accuracy.

[0096] In the present application, the unique genotype of the selected breed is selected. We believe that the descending combination of high-score sites cannot effectively serve as the basis for breed identification, but rather the dependent relationship between these sites should be distinguished, and these sites should be further classified for site selection from different categories. Therefore, the present application uses K-means clustering to select sites more suitable for breed identification of Saba pigs. Since a unique genotype is given, the genotype of the input sample can be simply compared to obtain the identification result when verifying the result. The training set of the present method basically contains pig breeds from all countries and regions, which can greatly increase the application breadth of the method.

[0097] We used the site selection method disclosed in the above article to select sites and verify the identification results of the SNP site data obtained in Example 1 of the present application:

[0098] First, the SNP site data obtained in Example 1 (SNP site data of 1016 samples) were subjected to Delta, Fst and In score calculation, sorted according to size, and the intersection of sites was obtained. The first 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100 sites of the intersection sites were selected to form 10 groups of site sets.

[0099] Secondly, LightGBM model training and evaluation are performed using the autogluon package of python: predictor = TabularPredictor(label=label).fit(train_data=df_fold_train, hyperparameters={'GBM':{}})

[0100] TabularPredictor(label=label).fit(train_data=df_fold_train, hyperparameters={'GBM':{}}). Among them, TabularPredictor is a model trainer in the autogluon package, label is the variety name of the training set, the fit method indicates the specific parameter setting: train_data=df_fold_train represents the training set of each fold data, and hyperparameters={'GBM':{}} indicates that LightGBM is set as the model used for training.

[0101] Thirdly, the accuracy is calculated using the Scikit-learn package of python with ten-fold cross-validation: StratifiedKFold(n_splits=10, shuffle=True, random_state=42). Among them, the StratifiedKFold class can ensure that the proportion of samples of each class in each fold is approximately the same as that of the complete set; n_splits=10 indicates that the data set is divided into 10 subsets, and in the subsequent cross-validation process, 9 subsets are used as training data and the remaining 1 subset is used as test data; shuffle=True indicates that the order of the data will be randomly shuffled before being divided; random_state=42 indicates a pseudo-random number generator seed, ensuring the repeatability of the code; the accuracy calculation formula is:

[0102]

[0103] Among them, A is the accuracy, TP represents the true positive, that is, the number of samples correctly predicted as Saba pigs; TN represents the true negative, that is, the number of samples correctly predicted as other varieties; FP represents the false positive, that is, the number of samples incorrectly predicted as Saba pigs; FN represents the false negative, that is, the number of samples incorrectly predicted as other varieties.

[0104] The results are shown in Table 5 and Figure 5

[0105] Table 5 Site number and accuracy results selected according to prior art method

[0106] Number of SNPs True positives True negatives False positives False negatives Accuracy 10 38 905 11 62 92.81％ 20 97 876 40 3 95.77％ 30 99 893 23 1 97.64％ 40 97 897 19 3 97.83％ 50 99 898 18 1 98.13％ 60 99 900 16 1 98.33％ 70 100 901 15 0 98.52％ 80 97 910 6 3 99.11％ 90 97 908 8 3 98.92％ 100 98 910 6 2 99.21％

[0107] Table 5 Site number and accuracy results selected according to prior art method Figure 5 ​As shown, the prior art site selection method needs to select at least 80 sites, and the accuracy can reach more than 99%, which is 62.5% more than the site selection of the method of the present application. In summary, the method of the present application can significantly reduce the number of sites required for identification, greatly improve the accuracy of identification, and produce significant benefits in guiding production practice.

[0108] It will be obvious to a person skilled in the art that, without departing from the spirit or essential characteristics of the application, the present application can be implemented in other specific forms. The present examples are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the specification as such. All changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0109] Furthermore, it should be understood that although the present specification describes only a single independent technical solution for each embodiment, the specification is described in this way only for the sake of clarity, and a person skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by a person skilled in the art.

Claims

1. A method for constructing a set of SNP loci for identification in Saba pigs, characterized in that, The method includes the following steps: S1. Calculate the breed score information of SNP loci of Saba pigs: Collect blood samples of Saba pigs for whole genome sequencing, and summarize them with the SNP data of various pig breeds in the IAnimal database to calculate the breed score of SNP loci of all pig breeds, namely the genetic difference Delta, the genetic differentiation index between paired populations Fst, and the allocation information In. S2. Select SNP loci from the above SNP loci that have unique alleles and have Delta > 0.49252, Fst > 0.14251, or In > 0.77959. Standardize the score of each SNP locus and classify them into three clusters using the K-means clustering method based on the standardized score of each SNP locus. In cluster 1, the Fst score is higher than the mean, and the Delta and In scores are higher than the mean. In cluster 2, the Fst score is higher than the mean, and the Delta and In scores are lower than the mean. In cluster 3, Fst scored below the mean, while Delta and In scored above the mean. S3. The SNP sites of the three clusters obtained in S2 are further selected using the following algorithm: (1) Calculate the common centroid of SNP sites in the three clusters. : ; in, As the central site, Let represent the coordinates of the i-th site in the j-th dimension, where n is the total number of sites and j=3; (2) Calculate the distance D between each SNP site and the center site in each cluster: ； (3) Iterative selection of sites: SNP sites are selected sequentially from cluster 1, cluster 2, and cluster 3 in descending order of D, and included in the site selection area in an ordered manner without replacement until all sites are selected. This process reorders the original sites, and the reordering order is the candidate site order for the SNP site set for Saba pig identification. S4. Reselection of candidate sites: Construct the LightGBM model, select the top 30 sites from the candidate sites of the above-mentioned SNP site set for Saba pigs, and use them as the identification SNP site set for Saba pigs.

2. The method for constructing a set of SNP loci for Saba pig identification according to claim 1, characterized in that, The K-means clustering used the R package NbClust, as follows: sc.km <- kmeans(sc_nl, 3, nstart=25), where sc.km is the final classification list; kmeans() is the K-means clustering function of the NbClust package; sc_nl is the scoring table of the SNP varieties; parameter 3 indicates clustering into three classes; and nstart=25 indicates generating 25 preset centroid values ​​to ensure stable classification results.

Images