Nonlinear kernels, dominance, and envirotyping data increase the accuracy of genome-based prediction in multi-environment trials

Abstract

Modern whole-genome prediction (WGP) frameworks that focus on multi-environment trials (MET) integrate large-scale genomics, phenomics, and envirotyping data. However, the more complex the statistical model, the longer the computational processing times, which do not always result in accuracy gains. We investigated the use of new kernel methods and modeling structures involving genomics and nongenomic sources of variation in two MET maize data sets. Five WGP models were considered, advancing in complexity from a main-effect additive model (A) to more complex structures, including dominance deviations (D), genotype × environment interaction (AE and DE), and the reaction-norm model using environmental covariables (W) and their interaction with A and D (AW + DW). A combination of those models built with three different kernel methods, Gaussian kernel (GK), Deep kernel (DK), and the benchmark genomic best linear-unbiased predictor (GBLUP/GB), was tested under three prediction scenarios: newly developed hybrids (CV1), sparse MET conditions (CV2), and new environments (CV0). GK and DK outperformed GB in prediction accuracy and reduction of computation time (~up to 20%) under all model–kernel scenarios. GK was more efficient in capturing the variation due to A + AE and D + DE effects and translated it into accuracy gains (~up to 85% compared with GB). DK provided more consistent predictions, even for more complex structures such as W + AW + DW. Our results suggest that DK and GK are more efficient in translating model complexity into accuracy, and more suitable for including dominance and reaction-norm effects in a biologically accurate and faster way.

The Modern Plant Breeding Triangle: Optimizing the Use of Genomics, Phenomics, and Enviromics Data

“…Thus, enviromics acts as a central bottleneck for the application of modern genomics-assisted prediction tools, especially for use across multiple environments. Novel approaches have integrated field trial data with DNA sequences using different sources of enviromics, such as linear and nonlinear reaction-norm models (e.g., Jarquín et al, 2014;Morais-Júnior et al, 2018;Millet et al, 2019;Monteverde et al, 2019;Costa-Neto et al, 2020a), crop growth model (CGM) outputs (Heslot et al, 2014;Rincent et al, 2017Rincent et al, , 2019, CGM integrated with GS (Cooper et al, 2016;Messina et al, 2018;Robert et al, 2020) and historical weather records to predict cultivars in years to come (de los Campos et al, 2020).…”

“…Another recent example is the approach that can increase the resolution in multi-environment prediction for stability by taking advantage of large-scale enviromics with different kernel methods (Costa-Neto et al, 2020a). The environmental relatedness among field trials can be shaped using linear covariances (as proposed by Jarquín et al, 2014) and nonlinear methods (Gaussian kernel, deep learning, and deep kernel) (Cuevas et al, 2016(Cuevas et al, , 2017(Cuevas et al, , 2018(Cuevas et al, , 2019Montesinos-López et al, 2018a,b, 2019b.…”

“…Progress toward the modernization of the statistical and quantitative genetic models for the analysis of plant breeding in multi-environment trials has become clearer as the availability of genomics, phenomics, and environments information has increased (see, among others, Vargas et al, 1998;Crossa et al, 2010;Burgueño et al, 2012;Heslot et al, 2014;Jarquín et al, 2014;Montesinos-López et al, 2017;Millet et al, 2019;Costa-Neto et al, 2020a;de los Campos et al, 2020;Robert et al, 2020). Thus, we see that all the elements described above offer a clear potential for the acceleration of genetic gains in plant breeding.…”

Modern Strategies to Assess and Breed Forest Tree Adaptation to Changing Climate

“…FIGURE 1 | Trans-disciplinary approaches (arrows) such as predictive breeding (GP) and machine learning (ML) promise supporting genome-wide marker-assisted (MAS) pre-breeding and breeding strategies for the selection of (A) "plus trees" in the wild, key (B) intra-and (C) inter-specific parental combinations, and (D) elite offspring from those parents. GP and ML should go beyond breeding and feedback (E) germplasm utilization and environmental niche classification (Cortés et al, 2013) and enviromics (Costa-Neto et al, 2020;Resende et al, 2020). Genomic-assisted characterizations, such as Genome-Wide Association Studies-GWAS (Neale and Savolainen, 2004), Genome-Environment Associations-GEA (Rellstab et al, 2015;Cortés and Blair, 2018;López-Hernández and Cortés, 2019) and Genome-Wide Selection Scans-GWSS (Zahn and Purnell, 2016), must also start considering more thoroughly (F) novel sources of local adaptation, (G) genetic-guided infusions and assisted gene flow (AGF), as well an overall systems genetics thinking (Ingvarsson et al, 2016;Myburg et al, 2019).…”

Data_Sheet_1.PDF

“…To expand our knowledge on the extent of the rootstock-scion interaction and speed up fruit tree breeding programs (Kumar et al, 2020;Peng et al, 2020;Santantonio and Robbins, 2020), further heritability estimates should be gathered on contrasting traits using multi-environment (Crossa et al, 2019;Costa-Neto et al, 2020) provenance ("common garden") and progeny trials with diverse panels of seedling and clonal rootstocks. The "genetic prediction" model used here to estimate pedigree-free heritabilities (Milner et al, 2000;Kruuk, 2004;Frentiu et al, 2008;Wilson et al, 2010;Berenos et al, 2014), or alternatively indirect genetic effect (IGE) models (Bijma, 2010(Bijma, , 2013Fisher and Mcadam, 2019), may be extended to field trials at a low genotyping cost, as few polymorphic SSR markers are enough to span the genetic relatedness gradient.…”