Identification And Analysis Of Genomic Variation In Chicken And Study On Genomic Selection In Layers

Genomic variations contribute greatly to the genetic and phenotypic diversity and are the main raw materials for both natural and artificial selection.Genomic variation includes single nucleotide polymorphism(SNP),insertion and deletion(INDEL)and structural variation(SV),and a comprehensive identification and understanding of these variation is essential for many genetic studies.To gain a better knowledge of INDEL variation in chicken genome,we applied next generation sequencing on 12 diverse chicken breeds.Over 1.3 million nonredundant short INDELs(1-49 bp)were obtained,the majority(95.76%)of which were less than 10 bp.The vast majority(92.48%)of our INDELs were novel and follow-up validation assays confirmed that most(88.00%)of the randomly selected INDELs represent true variations.Both the detected number and affected bases were larger for deletions than insertions.In total,INDELs covered 3.8 Mbp,corresponding to 0.36%of the chicken genome.The average genomic INDEL density was estimated as 0.49 per kb.INDELs were ubiquitous and distributed in a non-uniform fashion across chromosomes,with lower INDEL density in micro-chromosomes than in others,and some functional regions like exons and UTRs were prone to less INDELs than introns and intergenic regions.Nearly 620,253 INDELs fell in genic regions,1,765(0.28%)of which located in exons,spanning 1,358(7.56%)unique Ensembl genes.Many of them are associated with economically important traits and some are the homologues of human disease-related genes.The coding INDELs are valuable candidates for further elucidation of the association between genotypes and phenotypes.The chicken INDELs revealed by our study can be useful for future studies,including development of INDEL markers,construction of high density linkage map,INDEL arrays design,and hopefully,molecular breeding programs in chicken.Copy number variation(CNV)is a major source of genome polymorphism that directly contributes to phenotypic variation such as resistance to infectious diseases.Using next generation sequencing,we present a genome-wide assessment of CNVs that are potentially associated with genetic resistance to MD in two chicken lines that differ significantly in resistance to Marek’s disease(MD).Three chickens randomly selected from each line were sequenced to an average depth of 20×.Two popular software,CNVnator and Pindel,were used to call genomic CNVs separately.A total of 5,680 CNV regions(CNVRs)were identified after merging the two datasets,of which 1,546 and 1,866 were specific to the MD resistant or susceptible line,respectively.Over half of the line-specific CNVRs were shared by 2 or more chickens,reflecting the reduced diversity in both inbred lines.The CNVRs fixed in the susceptible lines were significantly enriched in genes involved in MAPK signaling pathway.We also found 67 CNVRs overlapping with 62 genes previously shown to be strong candidates of the underlying genes responsible for the susceptibility to MD.Our findings provide new insights into the genetic architecture of the two chicken lines and additional evidence that MAPK signaling pathway may play an important role in host response to MD virus infection.The rich source of line-specific CNVs is valuable for future disease-related association studies in the two chicken lines.Single step genomic prediction method has been proposed to improve the accuracy of genomic prediction by incorporating information of both genotyped and ungenotyped animals.The prediction performance of single step model,a two-step models and the pedigree-based models were compared in in a nuclear population of layers.A total of 1,341 chickens across 4 generations were genotyped by a 600K SNP chip.Four traits were analyzed,i.e.,body weight at 28 week(BW28),egg weight at 28 week(EW28),laying rate at 38 week(LR38),and Haugh unit at 36 week(HU36).In predicting offsprings,individuals from generation 1 to 3 were used as training data and 89 females from generation 4 were used as validation set.The performance of PBLUP,genomic BLUP(GBLUP),SSGBLUP and single step blending(SSBlending)were compared.On average,GBLUP performed better than PBLUP in LR38 but not in other traits,while the predictive ability of SSGBLUP and SSBlending were 14.6%and 11.8%higher than the PBLUP model.The prediction of the three genomic models was more biased than PBLUP models,except for trait LR38.It was concluded that single step models,especially SSGBLUP model,can yield more accurate prediction of genetic merits even with relatively small training size,and are preferable for practical implementation of genomic selection in layers.