Integrating environmental covariates and crop modeling into the genomic selection framework to predict genotype by environment interactions

Abstract

Development of models to predict genotype by environment interactions, in unobserved environments, using environmental covariates, a crop model and genomic selection. Application to a large winter wheat dataset. Genotype by environment interaction (G*E) is one of the key issues when analyzing phenotypes. The use of environment data to model G*E has long been a subject of interest but is limited by the same problems as those addressed by genomic selection methods: a large number of correlated predictors each explaining a small amount of the total variance. In addition, non-linear responses of genotypes to stresses are expected to further complicate the analysis. Using a crop model to derive stress covariates from daily weather data for predicted crop development stages, we propose an extension of the factorial regression model to genomic selection. This model is further extended to the marker level, enabling the modeling of quantitative trait loci (QTL) by environment interaction (Q*E), on a genome-wide scale. A newly developed ensemble method, soft rule fit, was used to improve this model and capture non-linear responses of QTL to stresses. The method is tested using a large winter wheat dataset, representative of the type of data available in a large-scale commercial breeding program. Accuracy in predicting genotype performance in unobserved environments for which weather data were available increased by 11.1% on average and the variability in prediction accuracy decreased by 10.8%. By leveraging agronomic knowledge and the large historical datasets generated by breeding programs, this new model provides insight into the genetic architecture of genotype by environment interactions and could predict genotype performance based on past and future weather scenarios.

Crop modelling: towards locally relevant and climate-informed adaptation

“…Vermeulen et al (2013) illustrate how crop-climate modelling and analyses can pre-emptively inform the magnitude of adaptation required across a range of time scales. Process-based crop models have also been applied to direct future research and crop breeding (Heslot et al 2014;Falloon et al 2015;Challinor et al 2016). The benefits of different adaptation strategies have often been compared by their relative impact on yield (e.g.…”

Needles: Toward Large-Scale Genomic Prediction with Marker-by-Environment Interaction

“…The paper by Heslot et al (2014) is the only one we found where M 3 E effects also were modeled explicitly and where an effort was made to detect QTL with specific environmental variations. However, a Bayesian Lasso approach was used, which made it computationally impractical to model all marker 3 environment combinations together with the marker effects at once, and thus a multistep approach was needed, leading to quite a complex procedure.…”

“…These early studies were mainly interested in mapping QTL that were stable across environments and thus obtained better estimates of the global genetic potential of an individual regardless of the environment. However, currently attention is directed more toward clarifying which QTL are susceptible to specific environmental covariables such as climatologic or soil conditions (Boer et al 2007;Heslot et al 2014). This makes it possible to define certain target populations of environments (TPEs) sharing some of the same environmental covariables so that environment-dependent QTL can be exploited for these TPEs (Heslot et al 2014).…”

“…However, currently attention is directed more toward clarifying which QTL are susceptible to specific environmental covariables such as climatologic or soil conditions (Boer et al 2007;Heslot et al 2014). This makes it possible to define certain target populations of environments (TPEs) sharing some of the same environmental covariables so that environment-dependent QTL can be exploited for these TPEs (Heslot et al 2014). Of course, the weather, playing an important factor in QTL expression (Boer et al 2007), is always partly unpredictable, so information about weatherdependent QTL can open up a whole new research field of balancing between yield optimization for a certain weather type and minimizing the risk of yield loss owing to bad weather predictions.…”

“…These intraenvironmental effects can be modeled together with a location effect and the G 3 E effects to immediately obtain the genotypic means across the environments in the first step of a two-step approach (Schulz-Streeck et al 2013b). However, in recent single-stage analyses, where the computation of genotypic means across environments is avoided and the interaction effects are explicitly modeled, the phenotypic records are mostly already corrected for spatial variations inside the environment (Burgueño et al 2012;Heslot et al 2014;Lopez-Cruz et al 2015). Nonetheless, the single-stage approach may include the modeling of these intraenvironmental effects to enable the direct analysis of the raw phenotypic data (Schulz-Streeck et al 2013a).…”