Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 3/10/2019
Publication Date: 6/23/2019
Citation: Wiggans, G.R., Van Raden, P.M., Null, D.J., Cole, J.B. 2019. Genomic predictions using more markers and gene tests. Journal of Dairy Science. 102(Suppl. 1):397(abstr. 462).
Technical Abstract: The number of markers used in US genomic predictions increased to 79,276 (or 80K) in December 2018 from the previous 60,671 (60K) used since 2014. The revised list includes more exact gene tests, removes poorer-quality markers, adds new variants from DNA sequence or high-density chips with larger effects on traits, and improves marker order using the new ARS-UCD1.2 reference map. Missing alleles were imputed for Holsteins by (1) imputing all bulls and their ancestors and (2) using those haplotypes as priors to impute the remaining 2 million females, which required 9 d to compute with 25 processors and 270 GB of memory. The 80K list increased computing times for other key programs by about 30%. Important variants now included directly (e.g., in DGAT1, ABCG2, ß-casein, and ß-lactoglobulin) had large effects on yield traits and the net merit index. Of the top 5 effects for each of 41 traits, only 34% were from the original 50,000-marker list for Holsteins, 55% for Jerseys, 53% for Brown Swiss, 33% for Ayrshires, and 38% for Guernseys. For Holsteins, a new sequence marker on chromosome 3 had the largest effect on final score, foot angle, feet-and-legs score, and rear legs (rear view). Gene tests for cholesterol deficiency, complex vertebral malformation, brachyspina, and calpain had large effects on somatic cell score, udder cleft, protein yield, and gestation length, respectively. For Jerseys, bGHR had a large effect on productive life. Genomic predictions improved more for breeds with larger reference populations. Individual predictions changed most for animals less related to the US population, with less complete pedigrees, or genotyped with lowest density chips. After excluding older cows genotyped using 3,000 markers and imputed dams, correlations of 80K with 60K predictions were about 0.99 for Jersey, Holstein, and Guernsey yield traits but higher for Ayrshires and Brown Swiss. Correlations for many other traits were lower. For Holsteins, correlations averaged a little less than 0.99 for the 6 new health traits and slightly less than 0.98 for type and calving traits. Reliability is expected to increase by 1 to 3 percentage points.