| Pork is one of the most important sources of animal protein for residents in china.Improving the quality of breeding pigs and selecting elite breeding pigs determines the potential for sustainable development of pig farming.From the historical point of view,conventional breeding methods have achieved great success in pig breeding,but "bottleneck effect" of these methods have also become increasingly serious.In recent years,the introduction and application of genomic selection has made up for the shortcomings of traditional breeding methods,especially in terms of decreased the generation interval and improved prediction accuracy of low heritability traits.More and more breeding companies have begun to use genomic selection to accelerate breeding progress.However,there are still many limiting factors in the current genomic selection,the most critical of which is the problem of prediction accuracy,which is closely related to different population structures,effective population size,genetic architecture of biological traits,and the density of markers.Compared with dairy cows,the characteristics of pig population structure,breed diversity,and economic traits are completely different.Only increasing the number of reference population cannot fundamentally solve the problem,and it is necessary to explore other effective technical to promote the application of genome selection.The development of genomic sequencing technology has made it possible for people to obtain whole-genome-level markers at low cost.Higher-density markers can obtain almost all linkage disequilibrium(LD)information,and even directly use causal variation as information for trait breeding value evaluation,which provides the possibility to solve population structure and complex genetic architecture of traits.In this study,the effects of different sequencing strategies on genome selection was investigated.First,a low-cost reduced-representation genome sequencing method was chosen and its effect on improving the accuracy of low heritability trait selection was evaluated in large white pigs.At the same time,by comparing with the SNP chip method,this study described the application value of RAD-seq technology.Second,extremely high-density markers were obtained from lowcoverage whole genome sequencing to evaluate genomic breeding values.Finally,the genetic basis of important economic traits such as birth weight and body length were analyzed using GWAS and fine mapping method,and multiple important candidate genes were identified.The detail of results is listed as follows:1.In this study,RAD-seq was used to genotype 618 large white pigs with average 580 Mb data per individual.Genotyping results showed that 79,725 high-quality SNPs can be obtained,with average call rate equal to 88.1% and average SNPs interval equal to 32.5 kb.To validate the efficiency of the RAD-seq on genomic selection,the 618 sows into two parts with 4:1 ratio were divided,which served as the reference and validation population,respectively.Then,the predictive accuracy and biasness of three methods including best linear unbiased prediction(BLUP),genomic BLUP(GBLUP)and single-step genomic BLUP(SSGBLUP)were studied,respectively,for validation population for which phenotypic records were masked.The results showed that GBLUP increased the prediction accuracy of breeding value from 0.109(TNB),0.067(NBA)and 0.009(LW)with BLUP to 0.220,0.184 and 0.205,respectively,and SSGBLUP performed similar prediction accuracy as GBLUP.Furthermore,GBLUP and SS-GBLUP improved the prediction biasness in comparison with BLUP.At the same time,through identification and utilization of large effect markers,compared with BLUP,the prediction accuracy has improved(14.1%-135.7%).2.In order to compare the effect of different genotyping methods in the same population on the accuracy estimation of genomic breeding values,RAD-seq and SNP chips were simultaneously carried out to genotype 453 large white pig populations.Under similar cost conditions,139,634 SNPs were obtained from RAD-seq with average 1.4G sequencing data per individual,while only 45,175 SNPs were obtained from 60 k SNP chip.Over 80% of SNPs obtained by RAD-seq have an average interval of less than 15 kb,and SNPs with similar interval in SNP chips is less than 20%.Comparative analysis showed that RADseq is twice as likely to find low-frequency SNPs(MAF<0.1)than SNP chips.GBLUP model was performed to estimate the genomic breeding value of growth,reproduction,body size,carcass traits.The results showed that except for 100 kg corrected day age and 100 kg corrected day weight gain,the estimation accuracy of other traits is not much different between the three treatments,including GBLUP-rad,GBLUP-chip and GBLUPrad&chip.For example,the difference in carcass traits is within 16.5%,the difference in body size traits is within 20%,and the difference in birth weight is within 18.5%.MBLUP analysis was performed using the large effect sites obtained by fine mapping of birth weight.The accuracy was improved by 42.3%-78.4% compared with other treatments,and the unbiasedness was closer to 1.3.Through low-depth resequencing analysis of 1097 large white pigs with different strategies,it was found that the mismatch rate,mapping rate,actual depth and coverage between different batches and different depths are in stability.In addition,correlation results showed that coverage is approximately 35% at 0.5×,approximately 60% at 1×,and approximately 80% at 2×.Compared with 10×genome sequencing data,the accuracy of genotype imputation is more than 99% for both 1× and 2×.At the same time,through comparing the 1×data with the data of the 60 K SNP chip,the results showed that the number of coincident SNPs can reach 67%,and the remaining SNPs are closely linked to the markers from 1×data.These results show that the 1×sequencing depth is the most suitable choice for cost effectiveness.4.A total of 15,506,511 SNPs obtained by low-depth resequencing(1×)were used to predict accuracy of genomic prediction in a population of 1097 large white pigs.Compared with conventional BLUP,ultra-high density markers can significantly improve the accuracy of genomic prediction of traits such as litter size,birth weight,and number of healthy piglets,and the range of improvement is 33.3%-128%.However,compared with the extracted 60 K SNP markers,the prediction accuracy of each trait genome is relatively small,ranging from 10.5% to 46.6%.5.The GWAS study was performed using the combined marker information obtained from the RAD-seq and SNP chips in Large white and Landrace pigs.Two regions that were significantly associated with pig birth weight were identified on chromosome 1 and chromosome 4,which could account for 6.36% and 4.25% of phenotypic variation,respectively.A region significantly associated with pig body length was identified based on data from RAD-seq,which explained 3.69% of the phenotypic variation.Six important candidate genes,including SKOR2,SMAD2,VAV3,NTNG1,TWIST2 and PER2 were obtained through fine mapping studies using breed-specific sequence-level marker imputation.These genes are related to important biological processes such as growth and development or bone development.Evaluation of RAD-seq and low-depth whole genome sequencing in pig genome selection could help promote genomic sequencing as the main genotyping methods for genome prediction and reduce dependence on foreign-made SNP chips.In the future,we can gradually accumulate massive amounts of whole-gene data to lay the foundation for building new breeding theories or assumptions in our country. |
PDF Full Text Request