Genetic and Genomic Resources in White Lupin and the Application of Genomic Selection

Abstract

Landraces represent extremely valuable and largely untapped genetic resources for white lupin improvement. The development of high-throughput, relatively low-cost genotyp-ing techniques, such as genotyping-by-sequencing (GBS), has allowed to develop dense genetic maps and to explore the application of genomic selection to predict breeding values of inbred lines or germplasm accessions for complex polygenic traits. We provide an unprecedented assessment of genomic selection in lupins, by assessing the ability of two selection models (Ridge regression BLUP and Bayesian Lasso) to predict grain yield and other traits of 83 landraces from nine historical cropping regions and eight varieties of white lupin that were autumn-sown in Northern Italy. GBS was applied to 3-4 genotypes per landrace and two genotypes per variety, analyzing cultivar allele frequencies for 6,578 polymor-phic SNP markers. The two selection models displayed similar predictive ability. Predictions proved highly accurate for grain yield, winter survival and onset of flowering, which displayed predictive abilities of 0.865, 0.852 and 0.838, respectively, based on cross-validation results. Moderately high predictive ability (0.626-0.495) emerged for pod fertility, individual seed weight, plant height, leaf size, and mainstem proportion of seeds and number of leaves. Genomic selection holds high promise for white lupin based on these results.