Integrating crop growth models with whole genome prediction through approximate Bayesian computation

Abstract

Genomic selection, enabled by whole genome prediction (WGP) methods, is revolutionizing plant breeding. Existing WGP methods have been shown to deliver accurate predictions in the most common settings, such as prediction of across environment performance for traits with additive gene effects. However, prediction of traits with non-additive gene effects and prediction of genotype by environment interaction (G×E), continues to be challenging. Previous attempts to increase prediction accuracy for these particularly difficult tasks employed prediction methods that are purely statistical in nature. Augmenting the statistical methods with biological knowledge has been largely overlooked thus far. Crop growth models (CGMs) attempt to represent the impact of functional relationships between plant physiology and the environment in the formation of yield and similar output traits of interest. Thus, they can explain the impact of G×E and certain types of non-additive gene effects on the expressed phenotype. Approximate Bayesian computation (ABC), a novel and powerful computational procedure, allows the incorporation of CGMs directly into the estimation of whole genome marker effects in WGP. Here we provide a proof of concept study for this novel approach and demonstrate its use with synthetic data sets. We show that this novel approach can be considerably more accurate than the benchmark WGP method GBLUP in predicting performance in environments represented in the estimation set as well as in previously unobserved environments for traits determined by non-additive gene effects. We conclude that this proof of concept demonstrates that using ABC for incorporating biological knowledge in the form of CGMs into WGP is a very promising and novel approach to improving prediction accuracy for some of the most challenging scenarios in plant breeding and applied genetics.

The quest for understanding phenotypic variation via integrated approaches in the field environment

“…using the model as an integrator that characterizes the growth environment), and second to extend the phenotyping per se at the level of the plot or plant, sometimes described as model-assisted phenotyping (Luquet et al, 2006;Rebolledo et al, 2015). A newly sought outcome from applications of CGMs is to combine their outputs of development and growth variables with knowledge from genomics and understanding of genetic architecture in order to predict the trait combinations, and eventually the allele combinations, that provide improved solutions in breeding (Technow et al, 2015).…”

3D sorghum reconstructions from depth images identify QTL regulating shoot architecture

“…Alleviating the phenotyping bottleneck for agriculturally important plants will help the world meet the increasing food and energy demands of the growing global population (Somerville et al, 2010;Alexandratos and Bruinsma, 2012;Cobb et al, 2013). Approaches to alleviate the plant phenotyping bottleneck fall into two broad categories: approaches that increase the number of individuals that can be grown and evaluated (Fahlgren et al, 2015b) and approaches that predict performance in silico to prioritize individuals to grow and evaluate (Hammer et al, 2010;Technow et al, 2015). Both of these approaches will be instrumental for increasing the rate of crop improvement, and both approaches are facilitated by advances in image-based phenotyping; multiple plant measurements can be acquired rapidly from images, and data from image-based phenotyping approaches also can inform performance prediction (Spalding and Miller, 2013;Pound et al, 2014).…”

Statistical methods for QTL mapping and genomic prediction of multiple traits and environments: case studies in pepper

“…Recently, efforts are being made to directly incorporate CGMs into the estimation of whole genome marker effects in GP using an Approximate Bayesian Computation (ABC) method (Technow et al, 2015). Technow et al, (2015) demonstrated the use of ABC as a mechanism for incorporating substantial biological knowledge embodied in the CGMs into a GP approach and showed that their proposed approach can be considerably more accurate than a benchmark GP method in predicting performance in environments represented in the estimation set as well as in previously unobserved environments for traits determined by non-additive gene effects.…”

“…Technow et al, (2015) demonstrated the use of ABC as a mechanism for incorporating substantial biological knowledge embodied in the CGMs into a GP approach and showed that their proposed approach can be considerably more accurate than a benchmark GP method in predicting performance in environments represented in the estimation set as well as in previously unobserved environments for traits determined by non-additive gene effects. A key difference in our approach to that of Technow et al, (2015) is the way in which component traits are introduced. We assume that all the components are observable / measurable for all the genotypes.…”

“…Also, a known relationship should exist between the complex trait and the component traits. Since CGMs contain explicit representations of development over time and integrate developmental and environmental information, they have the added advantage of being able to describe genotype-by-environment interactions (GEI) Technow et al, 2015;Chenu et al, 2009;Cooper et al, 2009). Therefore, component traits from one environment can be used to predict yield in another environment, provided the structure of the CGM is correct and GEI is small or absent for the component traits.…”