Genomic selection for lentil breeding: Empirical evidence

Abstract

Genomic selection (GS) is a marker-based selection initially suggested for livestock breeding and is being encouraged for crop breeding. Several statistical models are used to implement GS; however, none have been tested for use in lentil (Lens culinaris Medik.) breeding. This study was conducted to compare the accuracy of different GS models and prediction scenarios based on empirical data and to make recommendations for designing genomic selection strategies for lentil breeding. We evaluated nine single-trait (ST) models, two multiple-trait (MT) models, and a model that incorporates genotype × environment interaction (GEI) using populations from a lentil diversity panel and two recombinant inbred lines (RILs). The lines in all populations were phenotyped for five phenological traits and genotyped using a custom exome capture assay. Within-population, across-population, and across-environment genomic predictions were made. Prediction accuracy varied among the evaluated models, populations, prediction scenarios, and traits. Single-trait models showed similar accuracy in the absence of large effect quantitative trait loci (QTL) but BayesB outperformed all models when there were QTL with relatively large effects. Models that accounted for GEI and MT-GS models increased prediction accuracy for a low heritability trait by up to 66 and 14%, respectively. Moderate to high accuracies were obtained for within-population (range of .36-.85) and across-environment (range of .19-.89) predictions but across-population prediction accuracy was very low. Results suggest that GS can be implemented in lentil breeding to make predictions within populations and across environments, but across-population prediction should not be considered when the population size is small.

Genetic Analyses and Genomic Predictions of Root Rot Resistance in Common Bean Across Trials and Populations

“…Effective genomic cross-predictions of Pythium breeding values between the RR panel and the ADP do not support that the observed tolerance is exotic, but rather indicate a general quantitative tolerance to root rots available in Andean germplasm. Predictions across populations have been reported to be poor in lentil (Haile et al, 2020) and other crops (Charmet et al, 2014;Crossa et al, 2014), hence, the effective allelic diversity between ADP and RR panel must be similar. Taken together, our data suggests a quantitative mode of resistance of the traits evaluated here combining favorable different alleles from several precursors, in line with reports on Fusarium resistance (Ongom et al, 2012;Wang et al, 2018).…”

Deep Neural Networks for Genomic Prediction Do Not Estimate Marker Effects

“…The first two datasets we use for evaluation are published datasets of lentil ( Lens culinaris ) (21). The authors of this study examined GP performance within-population, across-populations, and across-environments using a variety of models.…”

“…The data used in the wheat and lentil experiments are available from the sources referenced by their authors (21, 22). The carinata data is available on request.…”

Genomics‐assisted lentil breeding: Current status and future strategies

“…For example, prediction accuracy of GEBV for grain yield was estimated to be 0.64 in soybean, indicating good potential of using GS for this low heritable trait in this crop (Jarquín et al, 2014). In lentil, different GS models and prediction scenarios have been evaluated for GS by including G × E interactions and multiple traits (Haile et al, 2020). In this study, GS was found useful to accelerate genetic gain within population and across environments (Haile et al, 2020).…”