P.M. VanRaden1, M.E. Tooker*1, K.M. Olson2, T.A. Cooper1, G.R. Wiggans1, and C.P. VanTassell3
1Animal Improvement Programs
Laboratory, ARS, USDA, Beltsville, MD
2National Association of
Animal Breeders, Columbia, MO
3Bovine Functional Genomics
Laboratory, ARS, USDA, Beltsville, MD
2011 J. Dairy Sci. (?)
© American Dairy Science Association,
2011. All rights reserved.
Individuals may download, store, or print single
copies solely for personal use.
Do not share personal accounts or passwords
for the purposes of disseminating this article.
Dairy cattle breeders have used a 50K chip since April 2008 and a less expensive, lower density (3K) chip since September 2010 in genomic selection. Evaluations from 3K are less reliable because genotype calls are less accurate and missing markers are imputed. After excluding genotypes with < 90% call rate and other edits, marker properties were compared for 8,305 animals with 3K, 54,643 with 50K version 1, 2,602 with 50K version 2, and 353 with a higher density (777K) chip. Animals were of 3 breeds, and all 4 chips were from Illumina. Numbers of markers selected were 2,614 from the 3K chip, 42,503 from 50K version 1, 41,019 from 50K version 2, and 632,665 from the 777K. Markers selected each had < 20% missing genotypes and < 2% parent-progeny conflicts, with proportionally stricter limits for markers with minor allele frequencies < 0.5. The selected 3K markers averaged 0.7% missing genotypes vs. 0.4% for the 50K markers. Parent-progeny conflicts averaged 0.07% for the selected 3K markers vs. 0.01% for the 50K markers. Properties of the 777K chip are similar to the 50K chips, but only 38,201 markers that match the 50K chip currently are used. Genomic evaluations were examined for 319 animals that had 3K genotypes in December and then 50K in February. Means were nearly identical (within 1 pound for fat and protein). The 3K evaluations had SD about 96% as large as 50K, in agreement with the 95% expected from the lower published reliability (64 vs. 71% for net merit). The correlations ranged from 0.92 to 0.96 across traits as compared to 0.46 to 0.66 for parent average with 50K. Version 2 of findhap.f90 will improve the 3K correlations to 0.94 to 0.97, improve average reliability by 2%, and improve the percentage of correctly imputed genotypes to 96.3% from 93.8% with version 1. Breeders can increase reliability affordably using lower density to impute higher density genotypes.
(Key words: Genomic evaluation, Imputation, Marker density)