Training set design in genomic prediction with multiple biparental families

Abstract

Genomic selection is a powerful tool to reduce the cycle length and enhance the genetic gain of complex traits in plant breeding. However, questions remain about the optimum design and composition of the training set. In this study, we used 944 soybean [Glycine max (L.) Merr.] recombinant inbred lines from eight families derived through a partial-diallel mating design among five parental lines. The cross-validated prediction accuracies for the six traits seed yield, 1,000-seed weight, protein yield, plant height, protein content, and oil content were high, ranging from 0.79 to 0.87. We investigated among-family predictions, making use of the special mating design with different degrees of relatedness among families. Generally, the prediction accuracy decreased from full-sibs to half-sib families to unrelated families. However, half-sib and unrelated families also showed substantial variation in their prediction accuracy for a given family, which appeared to be caused at least in part by the shared segregation of quantitative trait loci in both the training and prediction sets. Combining several half-sib families in composite training sets generally led to an increase in the prediction accuracy compared with the best family alone. The prediction accuracy increased with the size of the training set, but for comparable prediction accuracy, substantially more half-sibs were required than full-sibs. Collectively, our results highlight the potential of genomic selection for soybean breeding and, in a broader context, illustrate the importance of the targeted design of the training set. INTRODUCTIONGenomic selection was first suggested for animal breeding and has been shown to enhance the rate of genetic improvement Meuwissen et al., 2001). Genomic selection has also emerged as a valuable tool for plant breeding.

The performance of phenomic selection depends on the genetic architecture of the target trait

“…The genetic relatedness between the training set and the prediction set is a key factor for the success of genomic prediction in plant breeding (Brauner et al 2020 ; Zhu et al 2021 ). We therefore further investigated the predictive abilities of phenomic prediction and genomic prediction for grain yield among the three groups, i.e., among the diversity panel and the two DH populations (Fig.…”

“…The accuracy of genomic prediction is known to depend on the genetic relatedness of the individuals in the training and prediction sets (Lehermeier et al 2014 ; Würschum et al 2017 ; Brauner et al 2020 ; Zhu et al 2021 ). Predictions among the three groups revealed substantial differences between the phenomic and the genomic approach.…”

“…It jointly considers all markers in the entire genome and is thereby able to also capture the effects of small-effect loci (Meuwissen et al 2001 ; Heffner et al 2009 ). Over the last twenty years, genomic selection has proven to be a promising approach for the prediction of complex traits, as for example demonstrated for grain yield in various crops (Zhao et al 2012 , 2015 ; Würschum et al 2013 ; Pace et al 2015 ; He et al 2016 ; Thorwarth et al 2017 ; Zhu et al 2021 ). Nevertheless, in practical breeding programs, it is often challenging to obtain the required high-throughput genotypic data in each generation for a large number of newly established progeny, mainly due to the associated costs (Crossa et al 2017 ).…”

Phenomic selection is competitive with genomic selection for breeding of complex traits

“…The soybean population underlying this study, as well as QTL mapping and genomic prediction in this population, have been described previously (Kurasch et al, 2017;Zhu et al, 2020Zhu et al, , 2021aZhu et al, , 2021b. In brief, eight families were generated from the half-diallel mating design with five soybean cultivars adapted to Central Europe: 'Gallec' (P1), 'Primus' (P2), 'Protina' (P3), 'Sultana' (P4), and 'Sigalia' (P5).…”

“…For genomic prediction, relatedness between individuals in the training and prediction set is a major determinant for the prediction accuracy. Consequently, for the genomic approach there is a clear trend with highest predictive abilities being achieved within families (i.e., using full-sibs), a lower but usually acceptable predictive ability when half-sibs are used, and a very low predictive ability, often close to zero or even negative, when unrelated families are used for prediction (Brauner et al, 2020;Lehermeier et al, 2014;Riedelsheimer et al, 2013;Würschum et al, 2017;Zhu et al, 2021b). We therefore investigated the importance of relatedness for phenomic prediction.…”