Abstract #168
Section: Breeding and Genetics (orals)
Session: Breeding and Genetics II: Methodologies, Inbreeding and Breeding Strategies
Format: Oral
Day/Time: Monday 2:45 PM–3:00 PM
Location: Room 301 B
Session: Breeding and Genetics II: Methodologies, Inbreeding and Breeding Strategies
Format: Oral
Day/Time: Monday 2:45 PM–3:00 PM
Location: Room 301 B
# 168
Potential benefits from using a new reference map in genomic prediction.
Daniel J. Null*1, Paul M. VanRaden1, Derek M. Bickhart2, John B. Cole1, Jeff R. O'Connell3, Benjamin D. Rosen1, 1USDA Animal Genomics and Improvement Laboratory, Beltsville, MD, 2USDA Dairy Forage Research Center, Madison, WI, 3University of Maryland-Baltimore, Baltimore, MD.
Key Words: reference assembly, genomic prediction, liftover
Potential benefits from using a new reference map in genomic prediction.
Daniel J. Null*1, Paul M. VanRaden1, Derek M. Bickhart2, John B. Cole1, Jeff R. O'Connell3, Benjamin D. Rosen1, 1USDA Animal Genomics and Improvement Laboratory, Beltsville, MD, 2USDA Dairy Forage Research Center, Madison, WI, 3University of Maryland-Baltimore, Baltimore, MD.
Many genomic studies in cattle have used the 2009 reference assembly from the University of Maryland (UMD3.1). A new USDA Agricultural Research Service-University of California, Davis (ARS-UCD) assembly based on longer DNA reads from the same cow (Dominette) should improve sequence alignment, imputation, and genomic prediction. To test imputation, markers were converted from their previous map locations to new map locations in a process called liftover. Flanking sequences from array manifests or from the UMD3.1 map were remapped to the new assembly. For the 60,671 markers currently used in US genomic evaluations, > 99% aligned in forward direction to the same chromosome, but a few previously used markers were no longer usable. The new locations were then used to impute the 60,671 markers from genotype subsets available on most arrays for 1,748,453 Holsteins (HO), 215,800 Jerseys (JE), 32,724 Brown Swiss (BS), 4,834 Ayrshires (AY), and 3,517 Guernseys (GU). Average numbers of distinct haplotypes per segment decreased 5% for HO and from 1 to 30% for other breeds. Many previous problem areas no longer have excess numbers of haplotypes, particularly on the X chromosome and the pseudoautosomal region of X. Truly lethal haplotypes were more cleanly separated from false candidate haplotypes. Percentage of haplotypes with parent-progeny noninheritance dropped from 3.7 to 3.1 for HO, 4.4 to 3.9 for JE, 1.4 to 1.2 for BS, 2.5 to 1.5 for AY, and 1.6 to 1.4 for GU. Percentage of inherited haplotypes with 1 mistake dropped from 4.9 to 3.7 for HO, 4.3 to 3.6 for JE, 2.9 to 2.8 for BS, 3.5 to 2.8 for AY, and 3.0 to 2.7 for GU. Only a few segments such as on the left end of chromosome 8 had poorer properties for all breeds. Several regions of UMD3.1 were previously known to be on incorrect chromosomes and were excluded from use but can now be used with ARS-UCD. To test sequence alignment, paired-end reads from a HO bull were aligned to both maps, and 2.3% more paired reads aligned in the correct orientation within 5,000 base pairs. The new map improves genotype imputation, sequence alignment, and marker locations.
Key Words: reference assembly, genomic prediction, liftover
