Fig. 5
From: Limited haplotype diversity underlies polygenic trait architecture across 70 years of wheat breeding
Genetic architectures of 73 trait/year combinations (47 distinct phenotypes) as revealed by QTL mapping and genomic prediction. a Phenotypic variation explained by all genome-wide significant QTLs (thick bars) and by the full SNP-based genetic relationship matrix (heritability, thin bars and dots). Phenotypes measured in year 1 and year 2 are paired, shifted to the left and right, respectively. b LASSO prediction accuracy (Pearson correlation) across 50-fold cross validation; error bars show sds. c, d Prediction accuracy correlations (y-axis) vs sqrt (heritability) or QTL R2 (x-axis) in the test and training sets under ridge regression (rr) and LASSO genomic prediction models. Prediction into the test set is generally higher with LASSO, especially for traits where more variation is explained by genome-wide significant QTLs (d). e LASSO models usually include more SNPs when more heritable variation is unaccounted by genome-wide significant QTLs (x-axis is difference between heritability and QTL R2).