Abstract #M84
Section: Breeding and Genetics
Session: Breeding and Genetics: Application and methods in animal breeding - Swine, poultry, and other species
Format: Poster
Day/Time: Monday 7:30 AM–9:30 AM
Location: Gatlin Ballroom
Session: Breeding and Genetics: Application and methods in animal breeding - Swine, poultry, and other species
Format: Poster
Day/Time: Monday 7:30 AM–9:30 AM
Location: Gatlin Ballroom
# M84
Integration of haplotype analysis of functional genomic information with single SNP analysis improved accuracy of genomic prediction.
Cheng Tan*1,2, Dzianis Prakapenka1, Chunkao Wang1, Li Ma3, John R. Garbe4, Xiaoxiang Hu2, Yang Da1, 1Department of Animal Science, University of Minnesota, Saint Paul, Minnesota, 2State Key Laboratory of Agrobiotechnology, China Agricultural University, Beijing, China, 3Department of Animal and Avian Sciences, University of Maryland, College Park, MD, 4Minnesota Supercomputer Institute, University of Minnesota, Minneapolis, MN.
Key Words: haplotype, genomic selection, SNP
Integration of haplotype analysis of functional genomic information with single SNP analysis improved accuracy of genomic prediction.
Cheng Tan*1,2, Dzianis Prakapenka1, Chunkao Wang1, Li Ma3, John R. Garbe4, Xiaoxiang Hu2, Yang Da1, 1Department of Animal Science, University of Minnesota, Saint Paul, Minnesota, 2State Key Laboratory of Agrobiotechnology, China Agricultural University, Beijing, China, 3Department of Animal and Avian Sciences, University of Maryland, College Park, MD, 4Minnesota Supercomputer Institute, University of Minnesota, Minneapolis, MN.
We compared 3 methods of genomic prediction using additive and dominance effects of single SNPs and haplotype blocks. Method I was single SNP analysis of 423,131 SNPs covering all human autosomes from the Framingham Heart Study with over 6000 individuals. The next 2 methods add haplotype analysis of functional information to the single SNP analysis, i.e., Method II adds haplotype analysis of 595 “cholesterol-related genes” with 8,674 SNPs (2% of autosomes); Method III adds haplotype analysis of 9821 genes with 184,686 SNPs (36% of autosomes) after removing tiny genes without at least 2 SNPs. The results from 4 to 8 validation samples showed that adding haplotype analysis to single SNP analysis improved the prediction accuracy in most cases. Method II with cholesterol-related genes had the best prediction accuracy for total cholesterol with 4.78% increase in accuracy over single SNP analysis, and had stable accuracy increases across validation samples for all cholesterol phenotypes. Method III using all autosomal genes had the best accuracy for triglyceride with 17.75% increase in accuracy over single SNP analysis and tended to have the best performance across different phenotypes, but had larger variations than Method II across validation samples for cholesterol phenotypes. Results were also obtained from one validation sample for adding 3 other haplotype analyses to single SNP analysis: ChIPseq sites with 375,924 SNPs and average block size of 115.8Kb; non-hotspot blocks with each block between 2 crossover hotspots with 422,695 SNPs and average block size of 65Kb, and evenly divided blocks with block size of 100Kb of 422,814 SNPs. All 3 methods improved the prediction accuracy for most phenotypes but ChIPseq blocks mostly had better prediction accuracy than the other 2 methods, indicating that ChIPseq sites likely contained useful functional information not present in anonymous blocks. The results in this tudy tend to conclude that the integration of haplotype analysis of functional genomic information with single SNP analysis may improve the accuracy of genomic prediction for some phenotypes.
Key Words: haplotype, genomic selection, SNP
