Improved genomic prediction of clonal performance in sugarcane by exploiting non-additive genetic effects

Abstract

Non-additive genetic effects seem to play a substantial role in the expression of complex traits in sugarcane. Including non-additive effects in genomic prediction models significantly improves the prediction accuracy of clonal performance. In the recent decade, genetic progress has been slow in sugarcane. One reason might be that non-additive genetic effects contribute substantially to complex traits. Dense marker information provides the opportunity to exploit non-additive effects in genomic prediction. In this study, a series of genomic best linear unbiased prediction (GBLUP) models that account for additive and non-additive effects were assessed to improve the accuracy of clonal prediction. The reproducible kernel Hilbert space model, which captures non-additive genetic effects, was also tested. The models were compared using 3,006 genotyped elite clones measured for cane per hectare (TCH), commercial cane sugar (CCS), and Fibre content. Three forward prediction scenarios were considered to investigate the robustness of genomic prediction. By using a pseudo-diploid parameterization, we found significant non-additive effects that accounted for almost two-thirds of the total genetic variance for TCH. Average heterozygosity also had a major impact on TCH, indicating that directional dominance may be an important source of phenotypic variation for this trait. The extended-GBLUP model improved the prediction accuracies by at least 17% for TCH, but no improvement was observed for CCS and Fibre. Our results imply that non-additive genetic variance is important for complex traits in sugarcane, although further work is required to better understand the variance component partitioning in a highly polyploid context. Genomics-based breeding will likely benefit from exploiting non-additive genetic effects, especially in designing crossing schemes. These findings can help to improve clonal prediction, enabling a more accurate identification of variety candidates for the sugarcane industry.

Fig. 1

Accuracies of genomic prediction using a single kernel approach in a reproducible kernel Hilbert space (RKHS) model for a range of bandwidth parameters, h. This validation of h values was performed in order to select h values associated the highest prediction accuracy. A validation data set independent of the three forward prediction scenarios was chosen by using 1,320 clones from 2013 and 2014 as training population to predict 662 clones from 2015. TCH = tonnes of cane per hectare; CCS = commercial cane sugar; Fibre = Fibre content

Fig. 2

Genomic additive relationship matrix showing the proportion of genome shared amongst a total of 2,909 clones from 2013 to 2017 that were evaluated in final assessment trials of the Sugar Research Australia breeding program. The top and the side axis both represent the clones. Each coloured point represents the proportion of the genome each pair of clones have in common. Higher degrees of genomic relationships between clones are represented by a light colour (e.g. diagonal elements), while a pink shading represents a weaker genomic relationship.

Fig. 3

Decomposition of genetic variance into additive, dominance, additive–additive epistatic, and residual variance in two forward prediction scenarios. a Proportion of genetic variance in forward prediction scenarios 1 a/1b (1,825 clones from 2013–2015 used as training population) for six different covariance structures (see Table 2). b Proportion of genetic variance in forward prediction scenario 2 (2,397 clones from 2013–2016 used as training population) for six different covariance structures (see Table 2). Va = additive genetic variance; Vd = dominance genetic variance; Vaa = additive–additive epistasis variance; Ve = error variance; Model A = additive model; Model AH additive plus heterozygosity; Model AD additive plus dominance model; Model ADH additive, dominance plus heterozygosity; Model ADE  additive, dominance and epistatic effect; Model ADEH  additive, dominance, epistatic plus heterozygosity; TCH tonnes of cane per hectare; CCS   commercial cane sugar; Fibre = Fibre content

Fig. 4

Prediction accuracies for three key traits in different forward prediction scenarios measured as a Pearson’s correlation between genomic prediction and adjusted phenotypes of clones. Prediction accuracies for tonnes cane per hectare (TCH), commercial cane sugar (CCS), and Fibre content in, scenario 1a (1,825 clones from 2013–2015 used as training population to predict 739 clones from 2016); scenario 1b (1,825 clone from 2013–2015 used as training population to predict 691 clones from 2017); and scenario 2 (2,397 clones from 2013–2016 used as training population to predict 691 clones from 2017). Model A = additive model; Model AH additive plus heterozygosity; Model AD additive plus dominance model; Model ADH additive, dominance plus heterozygosity; Model ADE  additive, dominance and epistatic effect; Model ADEH  additive, dominance, epistatic plus heterozygosity. Error bars show the standard errors of the correlations between the genomic prediction and the adjusted phenotypes