-
Traita Training setb Fixed effectsc Mean predictive ability (r)d SDe Minimumf Maximumg Shoulder All families No 0.33 0.15 0.12 0.60 Honeycrisp-derived families 0.30 0.15 0.11 0.57 Minneiska-derived families 0.28 0.15 0.09 0.46 All families Yes 0.33 0.15 0.10 0.62 Honeycrisp-derived families 0.31 0.15 0.09 0.58 Minneiska-derived families 0.35 0.18 0.08 0.48 Lenticel All families No 0.34 0.10 0.23 0.53 Honeycrisp-derived families 0.30 0.11 0.11 0.47 Minneiska-derived families 0.35 0.10 0.26 0.46 a Shoulder russet ratings and the dual lenticel russet ratings; b Three training sets were used: (1) all breeding germplasm minus untested family, (2) all 'Honeycrisp'-derived families minus untested family, and (3) all 'Minneiska'-derived families minus untested family; c Fixed effect model included two SNPs at the previously reported LG2 and LG6 shoulder russet QTLs[8] as fixed effects; d Correlation between observed and predicted values using Pearson correlation averaged across testing set families; e Standard deviation of r all runs for each trait-training-model set; f Minimum r within each trait-training-model set; g Maximum r within each trait-training-model set. Table 1.
Summary of genomewide prediction for shoulder and dual lenticel russet ratings.
-
Model – Traita Testing setb Culling criteria Cull (n) Keep (n) Type A error (%)c Type B error (%)d GS – Shoulder Full-sib families Over 2.0 461 548 26.0 10.2 Over 2.5 248 761 15.0 10.8 Over 3.0 105 906 7.1 11.5 MAS – Shoulder < 2 LOW alleles 461 548 28.2 9.7 GS – Lenticel Over 2.5 885 124 22.6 7.1 Over 5.0 331 678 17.0 21.1 Over 7.5 13 996 0.9 14.4 GS – Shoulder Selections Over 2.0 87 76 – – Over 2.5 47 116 – – Over 3.0 17 146 – – MAS – Shoulder* < 2 LOW alleles 57 73 – – GS – Lenticel Over 2.5 149 14 – – Over 5.0 63 100 – – Over 7.5 13 150 – – a Model comparison between genomewide selection (GS) with all markers as random effects using the all training set and marker-assisted selection (MAS); targeted traits were shoulder russet formation and lenticel russet formation; b Germplasm used in the testing set; Breeding = all 13 'Honeycrisp'- and 'Minneiska'-derived families were used as a testing family (n = 1,009); Selections = all advanced selections (n = 163), * Only 130 individuals were assessed for MAS due to unresolved QTL alleles; c Type 1 error defined as individuals that were culled based upon predicted values but would have been kept based upon observed values; d Type 2 error defined as individuals that were kept based upon predicted values but would have been culled based upon observed values. Table 2.
Comparison of culling using genomewide selection (GS) at different thresholds and marker-assisted (MAS) for shoulder russet and genomewide selection at different thresholds for lenticel russet formation.
Figures
(0)
Tables
(2)