Sharing of either phenotypes or genetic variants can increase the accuracy of genomic prediction of feed efficiency

Table 7 Average weighted accuracies of GEBV for RFI and regression slopes of corrected phenotypes on GEBV (bias) of the four cohort and random fourfold cross-validation subsets

Training set	Number of traits^a	Variant set	Accuracy	Bias	Difference^b
Cohort
AUSc	1	50k	0.29	1.13
AUSc	1	HD	0.29	1.10	0.0
AUScAUSh	2	50k	0.25	0.90	− 3.5
AUScOVE	2	50k	0.52	1.31	22.9
AUScAUShOVE	3	50k	0.50	1.28	21.1
AUScOVE	1	50k	0.50	0.99	21.8
AUSc	1	50k + s-tr_G	0.34	1.28	5.6
AUSc	1	50k + m-tr_G	0.38	1.10	9.0
AUSc	1	50k + sm-tr_G	0.39	1.15	10.8
Random
AUSc	1	50k	0.32	1.07
AUSc	1	HD	0.32	1.14	0.5
AUScAUSh	2	50k	0.29	0.84	− 3.0
AUScOVE	2	50k	0.53	1.26	21.3
AUScAUShOVE	3	50k	0.50	1.17	18.4
AUScOVE	1	50k	0.52	0.93	20.1
AUSc	1	50k + s-tr_G	0.34	1.17	1.8
AUSc	1	50k + m-tr_G	0.37	1.10	5.5
AUSc	1	50k + sm-tr_G	0.38	1.15	6.6

Acronyms as described in Table 1
^aNumber of traits (datasets) used in each analysis
^bDifference in accuracy of GEBV (%) between using the Australian cow dataset (AUSc) with 50k GRM and the corresponding dataset

ISSN: 1297-9686