Strategy to imputation of large genomic deletions and utilizing them in mapping and genomic prediction in cattle

TY - ABST

T1 - Strategy to imputation of large genomic deletions and utilizing them in mapping and genomic prediction in cattle

AU - Mesbah-Uddin, Md

AU - Guldbrandtsen, Bernt

AU - Capitan, Aurélien

AU - Lund, Mogens Sandø

AU - Boichard, Didier

AU - Sahana, Goutam

N1 - Conference code: 6

PY - 2020

Y1 - 2020

N2 - Large genomic deletions are potential loss-of-function variants and can be lethal. Analyzing whole genome data of cattle, we report 8,480 large deletions (199bp to 773kb). Breakpoint analysis revealed that the majority of the deletions were most likely generated by micro homology-mediated end joining. Genotype likelihoods for deletions were computed using a Gaussian mixture model, assuming a linear relationship between observed read-depth with unobserved copy-number. Our strategy to impute deletion genotypes to the SNP array-typed population included extending the whole-genome sequence reference with deletion genotypes, followed by a two-step genotype refinement approach using Beagle4 and SHAPEIT2, and finally, joint imputation of SNPs, indels, and deletions using Minimac3. We achieved an imputation accuracy of r2>0.6 at MAF as low as 0.7% for SNPs and indels, and 0.2% for deletions. Subsequently, we performed GWAS for eight fertility traits. We report significant associations (P-value<5×10-8) of 30,384 SNPs, 178 indels and 3 deletions in Holstein, 17 SNPs in Jersey, and 23,481 SNPs, 189 indels and 13 deletions in Nordic Red cattle. Candidate genes within 500 kb of lead SNP included genes with annotations such as embryonic lethality, male and female infertility, oocyte degeneration, abnormal estrous cycle, and decreased ovulation rate in mouse, cattle, and zebrafish. We investigated the effect on genomic prediction of including one or several (genetic) variance components for imputed sequence variants and deletions in three cattle breeds. We used simulations to study the advantage of using across population information in genomic prediction, especially for population with small training data.

AB - Large genomic deletions are potential loss-of-function variants and can be lethal. Analyzing whole genome data of cattle, we report 8,480 large deletions (199bp to 773kb). Breakpoint analysis revealed that the majority of the deletions were most likely generated by micro homology-mediated end joining. Genotype likelihoods for deletions were computed using a Gaussian mixture model, assuming a linear relationship between observed read-depth with unobserved copy-number. Our strategy to impute deletion genotypes to the SNP array-typed population included extending the whole-genome sequence reference with deletion genotypes, followed by a two-step genotype refinement approach using Beagle4 and SHAPEIT2, and finally, joint imputation of SNPs, indels, and deletions using Minimac3. We achieved an imputation accuracy of r2>0.6 at MAF as low as 0.7% for SNPs and indels, and 0.2% for deletions. Subsequently, we performed GWAS for eight fertility traits. We report significant associations (P-value<5×10-8) of 30,384 SNPs, 178 indels and 3 deletions in Holstein, 17 SNPs in Jersey, and 23,481 SNPs, 189 indels and 13 deletions in Nordic Red cattle. Candidate genes within 500 kb of lead SNP included genes with annotations such as embryonic lethality, male and female infertility, oocyte degeneration, abnormal estrous cycle, and decreased ovulation rate in mouse, cattle, and zebrafish. We investigated the effect on genomic prediction of including one or several (genetic) variance components for imputed sequence variants and deletions in three cattle breeds. We used simulations to study the advantage of using across population information in genomic prediction, especially for population with small training data.

M3 - Conference abstract for conference

T2 - 6th International Conference of Quantitative Genetics

Y2 - 2 November 2020 through 12 November 2020

ER -