Skip to main content

Table 2 Prediction accuracies and bias of the GEBV

From: Efficient genomic prediction based on whole-genome sequence data using split-and-merge Bayesian variable selection

Subset

Ï€

Accuracy

Slope

SCS

PY

IFL

SCS

PY

IFL

50k

0.999

0.711

0.707

0.627

0.945

0.882

0.837

50k + 1060

0.999

0.712

0.702

0.603

0.909

0.843

0.737

50k + 5300

0.999

0.711

0.647

0.591

0.865

0.647

0.709

50k + 10,600

0.999

0.710

0.698

0.609

0.865

0.840

0.771

50k + 53,000

0.999

0.710

0.701

0.606

0.888

0.862

0.782

50k

0.900

0.714

0.703

0.625

0.953

0.879

0.839

50k + 1060

0.903

0.710

0.694

0.609

0.897

0.829

0.778

50k + 5300

0.912

0.704

0.685

0.607

0.862

0.813

0.772

50k + 10,600

0.921

0.705

0.691

0.601

0.867

0.831

0.764

50k + 53,000

0.956

0.706

0.689

0.601

0.884

0.847

0.773

  1. GEBV are computed using 50k SNPs alone or supplemented with 1060 to 53,000 sequence-based variants. Bias is assessed as the slope of the regression of observed EBV on the GEBV
  2. Analyses used either a π value of 0.999, or a value calculated assuming that 4154 (i.e. 0.1 % of 4,154,064) variants were assumed to have a large effect