Training Set Optimization for Sparse Phenotyping in Genomic Selection: A Conceptual Overview

Abstract

Genomic selection (GS) is becoming an essential tool in breeding programs due to its role in increasing genetic gain per unit time. The design of the training set (TRS) in GS is one of the key steps in the implementation of GS in plant and animal breeding programs mainly because (i) TRS optimization is critical for the efficiency and effectiveness of GS, (ii) breeders test genotypes in multi-year and multi-location trials to select the best-performing ones. In this framework, TRS optimization can help to decrease the number of genotypes to be tested and, therefore, reduce phenotyping cost and time, and (iii) we can obtain better prediction accuracies from optimally selected TRS than an arbitrary TRS. Here, we concentrate the efforts on reviewing the lessons learned from TRS optimization studies and their impact on crop breeding and discuss important features for the success of TRS optimization under different scenarios. In this article, we review the lessons learned from training population optimization in plants and the major challenges associated with the optimization of GS including population size, the relationship between training and test set (TS), update of TRS, and the use of different packages and algorithms for TRS implementation in GS. Finally, we describe general guidelines to improving the rate of genetic improvement by maximizing the use of the TRS optimization in the GS framework.

Comparison of linear and semi-parametric models incorporating genomic, pedigree, and associated loci information for the prediction of resistance to stripe rust in an Austrian winter wheat breeding program

“…Here, predictive ability for stripe rust resistance was higher under cross-validation than in the cross-year prediction frameworks, similar to previous reports where prediction accuracy for other traits was higher within populations than across populations (Thavamanikumar et al 2015 ; Haile et al 2021 ; Isidro y Sánchez and Akdemir 2021 ). Compared to a study on genomic prediction for stripe rust in bread wheat landraces from Afghanistan, we found similar levels of cross-validated predictive ability (Tehseen et al 2021 ), while our cross-validated genomic predictive ability results were lower than those reported in advanced lines from the CIMMYT bread wheat program (Juliana et al 2017 ) and in a panel of Central European winter wheat (Shahinnia et al 2022 ).…”

Designing optimal training sets for genomic prediction using adversarial validation with probit regression

“…For the selection of optimal training sets according with Isidro y Sánchez and Akdemir (2021) in the context of GS, the following two optimization processes can be performed: (a) targeted method when the testing set is defined a priori and the goal is to select an optimal training set from all available candidate individuals (training set) and (b) untargeted method when the goal is to select an optimal training set to predict the remaining individuals (that will conform the testing set), that is, now the testing set is not defined a priori (Akdemir et al, 2019, 2021; Akdemir & Isidro‐Sánchez, 2019). Therefore, the AV method that is proposed in this research is useful for the first type (type a) of optimization.…”

Alternative scoring methods of fusarium head blight resistance for genomic assisted breeding

“…The WheatSustain winter wheat panel is composed of 230 genotypes (cultivars and breeding lines) covering a wide genetic variability across Europe. We used the breeders’ knowledge and the mean of the coefficient of determination to selectthe training set lines for this experiment Laloë (1993) ; Isidro y Sánchez and Akdemir (2021) . The panel represents cultivars developed through breeding programs from Germany (157), Austria (50), Norway (14), Sweden (4), Denmark (3), Poland (1), and Switzerland (1).…”