1/1
Supplemental Material for Brauner et al., 2018
Version 3 2018-10-10, 16:23
Version 2 2018-10-05, 15:46
Version 1 2018-09-26, 16:39
dataset
posted on 2018-10-10, 16:23 authored by Pedro C. Brauner, Dominik Müller, Pascal Schopp, Juliane Böhm, Eva Bauer, Chris-Carolin Schön, Albrecht E. MelchingerSupplemental Tables
Table S1. List of abbreviations.
Table S2. Overview of the different scenarios used for genomic prediction (GP), brief description, mathematical description, validation scheme, sample size (Ni) and composition of the training set.
Table S3. Forecasted prediction accuracy (forecasted value ± standard error) using deterministic equations devised by Wientjes (ρW) and Daetwyler (ρD) for seven agronomic traits for scenario sL using Ni = 50 sampled from the elite flint (EF) lines, or doubled-haploid (DH) lines from the landraces GB, SF, SM, and WA, as well as the means across traits.
Table S4. Prediction accuracy (ρ ± standard error) of seven agronomic traits from genomic prediction (GP) for the scenario cLe obtained with GBLUP by simple validation (SV). The combination comprised Ni = 19 lines from each of the six doubled-haploid libraries (DHL) and the prediction set of each DHL comprised the remaining lines.
Supplemental Figures
Figure S1. Boxplots of seven agronomic traits for the elite flint (EF) lines and the doubled-haploid (DH) lines from landraces GB, SF, SM, WA, CG and RT.
Figure S2. Prediction accuracy (ρ) of seven agronomic traits from genomic prediction (GP) for the scenario LwL obtained with GBLUP by simple validation (SV) using the doubled-haploid (DH) lines from the landraces GB, SF, SM or WA as training set (Ni = 50) and DH lines from the other landraces as prediction set.
Figure S3. Boxplots of pairwise modified Rogers’ distances for the elite flint (EF) lines as well as for the doubled-haploid (DH) lines from landraces GB, SF, SM, WA, CG and RT.
Figure S4. Number of polymorphic markers determined with samples of Ni = 50 of the doubled-haploid libraries (DHL) and the elite flint (EF) lines, averaged over 30 repetitions, are shown on the diagonal. Modified Rogers’ distance between pairs of DHL are given above the diagonal and estimates of parameter θii* below the diagonal.
Figure S5. Histogram of pairwise coefficient of co-ancestry for the elite flint (EF) lines with pedigree information at least up to the grandparents.
Figure S6. Comparison of boxplots between leave-one-out cross-validation (LOOCV) and cross-validation (CV) for seven agronomic traits for the doubled-haploid (DH) lines from the landrace SM (Ni = 101) using incremental sample size. For each sample size, we plotted the prediction accuracy of 100 repetitions obtained for a random sample of genotypes in each run. The same genotypes were used for CV and LOOCV in each run. To obtain the same number of degrees of freedom for calculating the correlations between observed and predicted values in LOOCV and CV, the sample size of the prediction set in CV was chosen to be identical to the size of the training set.
Table S1. List of abbreviations.
Table S2. Overview of the different scenarios used for genomic prediction (GP), brief description, mathematical description, validation scheme, sample size (Ni) and composition of the training set.
Table S3. Forecasted prediction accuracy (forecasted value ± standard error) using deterministic equations devised by Wientjes (ρW) and Daetwyler (ρD) for seven agronomic traits for scenario sL using Ni = 50 sampled from the elite flint (EF) lines, or doubled-haploid (DH) lines from the landraces GB, SF, SM, and WA, as well as the means across traits.
Table S4. Prediction accuracy (ρ ± standard error) of seven agronomic traits from genomic prediction (GP) for the scenario cLe obtained with GBLUP by simple validation (SV). The combination comprised Ni = 19 lines from each of the six doubled-haploid libraries (DHL) and the prediction set of each DHL comprised the remaining lines.
Supplemental Figures
Figure S1. Boxplots of seven agronomic traits for the elite flint (EF) lines and the doubled-haploid (DH) lines from landraces GB, SF, SM, WA, CG and RT.
Figure S2. Prediction accuracy (ρ) of seven agronomic traits from genomic prediction (GP) for the scenario LwL obtained with GBLUP by simple validation (SV) using the doubled-haploid (DH) lines from the landraces GB, SF, SM or WA as training set (Ni = 50) and DH lines from the other landraces as prediction set.
Figure S3. Boxplots of pairwise modified Rogers’ distances for the elite flint (EF) lines as well as for the doubled-haploid (DH) lines from landraces GB, SF, SM, WA, CG and RT.
Figure S4. Number of polymorphic markers determined with samples of Ni = 50 of the doubled-haploid libraries (DHL) and the elite flint (EF) lines, averaged over 30 repetitions, are shown on the diagonal. Modified Rogers’ distance between pairs of DHL are given above the diagonal and estimates of parameter θii* below the diagonal.
Figure S5. Histogram of pairwise coefficient of co-ancestry for the elite flint (EF) lines with pedigree information at least up to the grandparents.
Figure S6. Comparison of boxplots between leave-one-out cross-validation (LOOCV) and cross-validation (CV) for seven agronomic traits for the doubled-haploid (DH) lines from the landrace SM (Ni = 101) using incremental sample size. For each sample size, we plotted the prediction accuracy of 100 repetitions obtained for a random sample of genotypes in each run. The same genotypes were used for CV and LOOCV in each run. To obtain the same number of degrees of freedom for calculating the correlations between observed and predicted values in LOOCV and CV, the sample size of the prediction set in CV was chosen to be identical to the size of the training set.
Supplemental Files
File S1. Phenotypic data with entry means for the locations Einbeck (EIN), Ludwigsburg (LUD), Eckartsweier (EWE) and Hohenheim (HOH).
File S2. Genotypic data for all 404 entries with 46,112 SNP markers.
