Development Of Genomic Selection Software Based On ICE Algorithm And Its Application In Genetic Analysis Of Pig Scrotal Hernia

Part?: Genomic selection is a new breeding method developed in the past ten years,the main principle is to assume that all QTLs are in LD with a marker or haplotype of markers,and the estimate breeding values of the traits can be calculated by the model.There are many models for genomic selection,the main difference between them are the prior assumptions for the SNP effect and calculation algorithm.The assumption of equal variance explained by all loci in the GBLUP model,which means most complex traits are controlled by very many polymorphisms with tiny effect,could be imprecise if a trait is affected by a small number of QTLs with large effects.In contrast,the prior assumption for the SNP effect of the Bayesmethod based on the MCMC algorithm are more flexible and realistic,except that the solution contains the MCMC algorithm,which requires tens of thousands of iterations to get a stable solution for each SNP,making it so time-consuming.Therefore,in this study,we have established a rapid genomic selection model based on the iterative conditional expectation algorithm with four zero-mean normal distributions as the prior assumption of the SNP effect,and the variance of which have been obtained accurately based on the 374 standardized phenotypes in a F2 population.In order to verify the reliability and calculation speed of our model,we randomly gave the ratio of variance in the prior assumption,and selected two traits for accuracy verification,the result showed that the original model the better accuracy,indicating that the estimated variance ratio in this study has a high reliability and compatibility.Then,the four software of GBLUP,BayesR,BayesA,BayesB were selected for comparison.In addition,we have compared the prediction accuracy and computation time for 6 traits with different heritability and different genetic architecture in these five softwares using cross-validation in order to explore the differences in genomic selection of the traits with different heritability and different genetic architecture.Our results showed that the prediction accuracy of BayesR,BayesA and BayesB performs best when traits were controlled by several major QTLs and explained a large proportion of phenotypic variance in some SNPs,while FCF-MixP performs better than GBLUP.The prediction accuracy of these five methods was similar when traits follows polygenes model,and FCF-MixP is slightly better than others.Next,we performed FCFMixP in a duroc population to test its performance in other groups,the results showed that the prediction accuracies were 0.9056,0.8176,0.9905,0.8612 and 0.8892 for FCF-MixP,GBLUP,BayesR,BayesA and BayesB,respectively.Finally,we counted the calculation time of these five softwares and explored the change of computing time for FCF-MixP with the increase of the number of individuals in the reference group.The computational speeds for ICE-based FCF-MixP and GBLUP are on the same order of magnitude when there are fewer individuals in the reserence group,which are much faster than those of MCMC-based Bayesian methods,and the calculation time of FCF-MixP increases linearly with the increase of individuals in the reference group.With the increasing of genomic data and the individuals in the reference group,the genomic selection method based on MCMC algorithm will face a serious computational burden,and the inverse process of the kinship matrix in direct method will also become very difficult.As a result,FCF-MixP,which combines accuracy and computational efficiency,will be widely used the future.Our FCF-MixP method can be freely accessed at https://github.com/xuwenwu24/FCF-MixP.git.Part 2: Pig scrotal hernia is one of the most common congenital defect triggered by both genetic and environmental factors,leading to severe economic loss as well as poor animal welfare in the pig industry.Identification and implementation of genomic regions controlling scrotal hernia in breeding is of great appeal to reduce incidences of hernia in pig production.This study designed a special scrotal hernia group by the mating of the health full-sibs of the individuals,which included 246 individuals,of whom 18 were affected,and the prevalence rate is about three times of random mating groups.This special designed population has verified the hereditability of scrotal hernia and had a higher rate of affected individuals,which enhanced the effectiveness of genetic analysis.Next,we performed genome-wide association study(GWAS)in this Specially Designed Population in two scenarios,i.e.,the target panel data before and after imputation.The result showed that a 24.8Mb region(114.1-138.9Mb)on chromosome 8 was found to be significantly associated with pig scrotal hernia.To further narrow the region,two methods including genetic differentiation analysis and Linkage disequilibrium and linkage analysis(LDLA)were appropriate to provide for further analysis,and haplotype sharing analysis was carried out to identify the potential gene underlying the QTL.the result of genetic differentiation analysis showed a strong genetic differentiation signal between 116.1 and 132.7Mb on SSC8,and the result of the LDLA analysis showed that the most likely confidence interval is in the 121-123.99 Mb of chromosome 8.In addition,the QTL interval was refined to 2.99 Mb by combining LDLA and genetic differentiation analysis.Finally,two susceptibility haplotypes were identified through haplotype sharing analysis,with one potential causal gene(EIF4E)in it.Our study provided deeper insights into the genetic architecture of pig scrotal hernia and contributed to the disease-resistant breeding of scrotal hernia,and provided some reference for the study of human scrotal hernia.