A genotype filling method for yak genome sequencing data

Abstract

The present invention discloses a genotype filling method for yak genome sequencing data, which extracts molecular markers according to the deep genome sequencing data of yaks, converts the format of low-density resequencing molecular marker data, and classifies and groups yak populations based on cluster analysis; uses the genetic relationship between individuals within each classification group and the estimated genetic parameters, establishes a genotype filling model based on a hidden Markov model, estimates the best possible genotype of the missing value, and performs macro prediction for a small population; and finally sorts out the genotype data after the genome of the entire large population of yaks is filled. The present invention utilizes the background that the genetic structure of a small population of yaks is stronger than the overall genetic relationship, combines the hidden Markov prediction model and the round-by-round filling technology to improve the accuracy of yak genome sequencing data filling.

Description

Technical Field

[0001] The invention relates to the technical field of genome information filling, in particular to a method for filling genotypes in yak genome sequencing data. Background Art

[0002] Whole Genome Sequencing (WGS) is the process of sequencing all the genomes of a cell or organism to obtain complete genome information. Whole genome sequencing data can find the connection between genes and phenotypes, and is of great significance in mining functional genes related to important economic traits of plants and animals and analyzing genetic mechanisms.

[0003] Yaks in my country are mainly distributed in alpine grasslands and alpine canyon areas above 3,000 meters above sea level on the Qinghai-Tibet Plateau. They have excellent traits of high cold and low oxygen tolerance and are an important means of production that herders rely on. Due to long-term overbreeding and inbreeding caused by traditional yak breeding methods, the reproductive capacity, growth rate, adult size and milk production of domestic yaks have declined, affecting the development of local animal husbandry. By analyzing the whole genome sequencing data, candidate genes related to growth traits can be found, but the whole genome sequencing data of yaks generally have many missing values, and the missing values ​​need to be filled using the genotype filling method.

[0004] The basic principle of the genotype filling method is to construct the haplotype information shared between the reference group and the target group based on the genotype information provided by the reference group, compare the haplotype information shared between the target group and the reference group, fill in the missing genotype information of the target group, and obtain complete genotype data. Genotype filling methods can be roughly divided into two categories, one is computationally intensive, such as IMPUTE, MACH and fastPHASE, and the other is computationally efficient, such as PLINK, MINIMAC and BEAGLE. The computationally intensive genotype filling software fully considers all known genotype information during the filling process, making the estimation of unknown genotype information more accurate, but the filling takes a long time. The computationally efficient genotype filling software only focuses on the genotype information of a part of the markers adjacent to the specific SNP site during the filling process, and the filling time is shorter, but the filling accuracy is relatively lower. In the study of Japanese black cattle, the accuracy of genotype calculation of the Japanese black cattle population was evaluated from three aspects using BEAGLE software. For the effect of different numbers of missing genotypes, the filling match rate and correlation were higher for 50K, 26K, and 20K, but lower for 7K. In the beef cattle study, the software BEAGLE, FIMPUTE, and IMPUTE2 were used to fill in a multi-breed beef cattle population. The filling match rates for purebred populations ranged from 94.20% to 97.93% for FIMPUTE software, 95.35% to 98.31% for IMPUTE2 software, and 90.02% to 96.38% for BEAGLE software. The filling match rates for crossbred animals were 54.15-97.53% (FIMPUTE), 57.04-97.46% (IMPUTE2), and 54.35-95.64% (BEAGLE). In the study of cattle, three reference genomes were selected, and after genotyping the data, PEDIMPUTE, FINCHAP, FIMPUTE and BEAGLE software were used for filling. The filling matching rate of FIMPUTE software was the highest, about 95%, followed by BEAGLE software, about 92%. The filling matching rates of the other two software varied greatly due to the influence of genotyping. Genotype filling technology is widely used in humans, chickens, sheep, pigs, and plants.

[0005] However, yaks have unique genetic structure, population composition and genome complexity, which makes it more difficult to accurately fill in yak sequencing data; the genotype filling software or methods in the existing technology are not suitable for yak genotype filling. Therefore, it is urgent to develop a genotype filling method for yak genome sequencing data. Summary of the invention

[0006] The purpose of the present invention is to solve the deficiencies in the prior art and provide a method for filling genotypes in yak genome sequencing data.

[0007] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0008] A method for filling genotypes in yak genome sequencing data comprises the following steps:

[0009] 1) Cluster and classify yak populations based on existing yak genotype data, and use the classified populations as the basic populations;

[0010] 2) Using the hidden Markov model and the genetic structure relationship and genetic parameters within the population, the probability of unknown genotypes within each classification population is estimated, and the genotype with the greatest possibility is finally used as the final prediction result;

[0011] 3) Integrate the genotype data of each classification group. If there is still unknown genotype data, continue to repeat steps 1) to 2) until the genotype data of all classification groups are fully filled.

[0012] Furthermore, the step 1) specifically includes: converting the genotype polymorphism data SNP of the yak genotype data into Hapmap, bed or numeric format, and using the existing K-means and Cluster clustering methods to comprehensively determine the specific classification group of the yak individual, that is, the two clustering methods simultaneously determine that the individual belongs to the same group, otherwise the individual is classified into a separate group.

[0013] Furthermore, the step 2) specifically includes:

[0014] (1) The genetic structure of individuals in a taxonomic group is estimated using a molecular genetic relationship matrix. The degree of proximity between individuals determines the initial confidence probability of the genotype estimation between individuals. Genetic parameter evaluation mainly refers to the determination of linkage disequilibrium relationships, i.e., haplotype estimation, which is used to determine the distance between genes in a taxonomic group, i.e., the length of the credible haplotype.

[0015] (2) using a hidden Markov model (HMM) to calculate the probability of one marker to the next in the haplotype of a reference panel, where the reference panel refers to multiple individuals in the taxonomic group with known genotypes at locations where the genotypes of other individuals are unknown;

[0016] (3) At each marker, the sum of the probabilities of the allele markers is used as the estimated probability of the allele; the sum of the probabilities from the first marker to the last marker is the specific haplotype probability;

[0017] (4) A model is then established based on the shared gene sequences between the target sample and the reference panel, and the calculated probabilities of the allele markers in the reference panel are used to predict the maximum likelihood of various genotypes at the same markers in the target sample.

[0018] Compared with the prior art, the genotype filling of yak genome sequencing data using the method of the present invention can improve its filling accuracy, has a good filling effect, and takes a short filling time, and can quickly obtain complete yak genome information. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The accuracy of the three filling methods under different missing rate conditions.

[0020] Figure 2 The correlation between the three filling methods under different missing rate conditions.

[0021] Figure 3 The filling time of the three filling methods under different missing rate conditions. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the invention.

[0023] This embodiment provides a method for filling genotypes in yak genome sequencing data, comprising the following steps:

[0024] This example uses the genome sequencing data of 354 yaks as demonstration data. The original data can be downloaded from the website ( https: / / www.animalgenome.org / repository / pub / NWAU2019.0430 / ), which contains three types of files, namely ped files, map files and csv files. The ped files contain information on 98,688 SNP sites of 354 yaks, the map files contain genetic map information of 354 yaks, and the csv files contain four phenotypes of weight, shoulder height, body length and chest circumference of 354 yaks; the ped files are converted into Hapmap format genome information files. 10k data are randomly selected from all yak genome data to generate new files as test data for this experiment. Each repetition will randomly select 10,000 SNP sites again, and a total of 50 repetitions will be performed to eliminate the interference caused by random sampling. Three genotype filling methods are used for filling under the conditions of missing rates of 5%, 10%, 15% and 20%, respectively.

[0025] 2) Cluster and classify the 354 yaks according to the existing genotype data, and use the classified groups as the basic groups:

[0026] The K-means and Cluster algorithms were used to estimate the population genetic distance of 354 yaks. Individuals with close genetic distances were grouped together based on cluster analysis. Since the randomly selected SNPs were different for each repetition, the number of optimal clustering groups in each repetition was also different. We filled in the missing values ​​for each repetition and finally calculated the average filling accuracy under the total number of repetitions.

[0027] 3) Using the hidden Markov model and the genetic structure relationship and genetic parameters within the population, the probability of unknown genotypes within each classification group is estimated, and the genotype with the greatest probability is used as the final prediction result. For small classification groups (when the number of individuals is less than 5), the entire yak population is used as the reference group for estimation and prediction;

[0028] (1) The genetic structure of individuals in a taxonomic group is estimated using a molecular genetic relationship matrix. The distance between individuals determines the initial confidence probability of the genotype estimation between individuals. The genetic parameter evaluation mainly refers to the determination of linkage disequilibrium relationships (i.e., haplotype estimation), which is used to determine the distance between the credible estimated distances of genes in the taxonomic group (i.e., the credible haplotype length).

[0029] (2) using a hidden Markov model (HMM) to calculate the probability of one marker to the next in the haplotype of a reference panel, where the reference panel refers to multiple individuals in the taxonomic group with known genotypes at locations where the genotypes of other individuals are unknown;

[0030] (3) At each marker, the sum of the probabilities of the allele markers is used as the estimated probability of the allele; the sum of the probabilities from the first marker to the last marker is the specific haplotype probability;

[0031] (4) A model is then established based on the shared gene sequences between the target sample and the reference panel, and the maximum likelihood of various genotypes at the same marker in the target sample is predicted using the calculated probabilities of the allele markers in the reference panel;

[0032] (5) For those taxa without multiple individuals (less than 5 individuals), since the number of individuals is too small to provide strong support for the reference panel, we will use the entire large population for estimation. The estimation and prediction process is similar to (1)-(4).

[0033] 4) Integrate the genotype data of each classification group. If there is still unknown genotype data, continue to repeat steps 2) to 4) until the genotype data of all groups are fully filled.

[0034] Furthermore, in order to verify the superiority of the genotype filling method of the present invention, the sequencing data of yak were filled using the written StochasticImpute function, the impute.knn algorithm and the three genotype filling methods of the present invention, and the filling effects of the three filling methods under different missing rate conditions were explored, and the filling accuracy, correlation and filling time were used as evaluation indicators for evaluation.

[0035] The filling principle of the StochasticImpute function is to use the frequency of alleles to fill missing values, that is, taking all individuals as the reference group, calculating the frequency of all alleles in each SNP locus, and filling the missing values ​​in the SNP locus with the allele with the highest frequency. According to the numerical genotype file, the StochasticImpute function is used to calculate the frequencies of the three genotypes 0, 1, and 2 respectively, and the allele with the highest frequency is used as the filling type.

[0036] The impute.knn algorithm requires calling the "impute" package in the R language. This algorithm uses the genotype values ​​of a specific number of neighbors in the genotype data to fill in the genotypes of individuals with missing values. First, the candidate neighbors need to be classified, and then the distance formula is used to calculate the distance between the gene with missing values ​​and the candidate neighbors. The gene coordinates used to calculate the distance should be the non-missing elements in the gene. For the case where the candidate neighbors may lack coordinates for calculating the distance, the average value of the non-missing elements needs to be calculated. After finding the k nearest neighbors of a gene, the genotype with missing values ​​is determined based on the category to which most of the neighbors in these k nearest neighbors belong, and then the missing elements are estimated by averaging their adjacent non-missing elements. The k value set in this article is 3, and Euclidean distance calculation is selected to calculate the distance between neighboring data.

[0037] In order to compare the filling effects of the three genotype filling methods, this paper will use filling accuracy, correlation and filling time as evaluation criteria. The filling accuracy refers to the ratio of the number of correctly filled genotypes to the number of genotypes that need to be filled. match is the number of accurately predicted genotypes after filling, N total is the number of all missing genotypes.

[0038]

[0039] The correlation is the correlation between the true genotype and the imputed genotype. impute is the genotype after filling, G real is the true genotype.

[0040]

[0041] The filling time refers to the time from the start of filling to the end of filling. Use R language system.time to mark the start and end time, and finally take the difference to calculate the accurate calculation time.

[0042] For genotype data, we randomly set missing rates of 5%, 10%, 15%, and 20%, and used three genotype filling methods to fill in the missing data. This was repeated 30 times to obtain the filling accuracy, as follows: Figure 1 .

[0043] Depend on Figure 1 It can be seen that under different missing rate conditions, the filling accuracy of the present invention is the highest, and the filling accuracy rates are 0.8630, 0.8617, 0.8609 and 0.8599 respectively, followed by the impute.knn algorithm, with filling accuracy rates of 0.8125, 0.8123, 0.8125 and 0.8125 respectively, and the lowest filling accuracy is the StochasticImpute function, with filling accuracy rates of 0.6265, 0.6214, 0.6148 and 0.6063 respectively. As the missing rate increases, the filling accuracy rates of the present invention and the impute.knn algorithm are reduced, while the filling accuracy rate of the StochasticImpute function is basically unchanged.

[0044] While calculating the filling accuracy, calculate the correlation and get the correlation of the three filling methods repeated 30 times, take the average value and get Figure 2 .

[0045] Depend on Figure 2 It can be seen that under different missing rate conditions, the correlation between the StochasticImpute function and the impute.knn algorithm is relatively high. The correlations of the StochasticImpute function are 0.2216, 0.2190, 0.2193, and 0.2193, respectively, while the correlations of the impute.knn algorithm are 0.2218, 0.2204, 0.2185, and 0.2197, respectively. The correlations of the present invention are 0.2204, 0.2026, 0.1831, and 0.1637, respectively. As the missing rate increases, the correlation of the present invention gradually decreases, and is almost lower than the correlation of the other two filling methods. As the missing rate increases, the correlation between the StochasticImpute function and the impute.knn algorithm does not change substantially.

[0046] When using the three filling methods, record the filling time obtained by repeating the three filling methods 30 times, take the average value, and standardize it with a logarithmic function to obtain Figure 3 .

[0047] Depend on Figure 3It can be seen that the filling time of the present invention is the longest, and the filling time is 380.7 seconds, 465.3 seconds, 531.0 seconds and 604.5 seconds respectively, followed by the impute.knn algorithm, which takes 13.0 seconds, 14.0 seconds, 14.4 seconds and 15.4 seconds respectively, and the StochasticImpute function takes the least time to fill, which takes 7.2 seconds, 7.7 seconds, 7.9 seconds and 8.0 seconds. As the missing rate increases, the filling time of the three filling methods is increasing, among which the filling time of the present invention increases the most, and the filling time is 604.5 seconds under the condition of 20% missing rate.

[0048] Summary: Through the comparison of the three filling methods, the filling matching rate of the present invention is higher. The 5% improvement means that if there are 100,000 missing genotypes, the present invention will improve the accurate prediction of 5,000 genes compared with the other two methods; the correlation of the impute.knn algorithm is higher. The disadvantage of the present invention in correlation may be due to the fact that only three genotypes are designed, which statistically causes bias in correlation calculation; the filling time of the StochasticImpute function is shorter, but the calculation time of the present invention is within an acceptable range. When the missing rate is less than 20%, the filling effect of the genotype filling method of the present invention is better. It is recommended to use a missing rate of 20% to filter the sequencing data. The filtered data can be filled with the genotype filling technology of the present invention to obtain genotype data with higher credibility.

[0049] The technical solution of the present invention is not limited to the above-mentioned specific embodiments. All technical variations made according to the technical solution of the present invention fall within the protection scope of the present invention.

Claims

1. A method for filling genotypes in yak genome sequencing data, characterized in that: The following steps are involved: Step 1) Cluster and classify the yak population based on the existing yak genotype data, and use the classified population as the basic population; Step 2) Using the hidden Markov model and the genetic structure relationship and genetic parameters within the population, the probability of unknown genotypes within each classification population is estimated, and the genotype with the greatest probability is finally used as the final prediction result; The step 2) specifically includes: (1) The genetic structure of individuals in a taxonomic group is estimated using a molecular genetic relationship matrix. The degree of proximity between individuals determines the initial confidence probability of the genotype estimation between individuals. Genetic parameter evaluation refers to the determination of linkage disequilibrium relationships, i.e., haplotype estimation, which is used to determine the distance between genes in a taxonomic group, i.e., the length of the credible haplotype. (2) Using a hidden Markov model (HMM) to calculate the probability of one marker to the next in the haplotype of a reference panel, where the reference panel refers to multiple individuals in the taxonomic group with known genotypes at locations where the genotypes of other individuals are unknown; (3) At each marker, the sum of the probabilities of the allele markers is used as the estimated probability of the allele; the sum of the probabilities from the first marker to the last marker is the specific haplotype probability; (4) A model is then established based on the shared gene sequences between the target sample and the reference panel, and the calculated probabilities of the allele markers in the reference panel are used to predict the maximum likelihood of various genotypes at the same marker in the target sample; Step 3) Integrate the genotype data of each classification group. If there is still unknown genotype data, continue to repeat steps 1) to 2) until the genotype data of all classification groups are fully filled.

2. The method for filling genotypes of yak genome sequencing data according to claim 1, characterized in that: The step 1) specifically includes: converting the genotype polymorphism data SNP of the yak genotype data into Hapmap, bed or numeric format, and using the existing K-means and Cluster clustering methods to comprehensively determine the specific classification group of the yak individual, that is, the two clustering methods simultaneously determine that the individual belongs to the same group, otherwise the individual is classified into a separate group.

Images