Integrating Crop Growth Models with Whole Genome Prediction through Approximate Bayesian Computation

Area of Science:

• Plant breeding
• Genetics
• Computational biology

Background:

• Genomic selection (GS) and whole genome prediction (WGP) are revolutionizing plant breeding, but struggle with complex traits and genotype by environment interactions (G×E).
• Current WGP methods primarily rely on statistical approaches, largely overlooking the integration of biological knowledge.
• Predicting traits influenced by non-additive gene effects and G×E remains a significant challenge in plant breeding.

Purpose of the Study:

• To introduce and validate a novel WGP approach that integrates biological knowledge using crop growth models (CGMs) and approximate Bayesian computation (ABC).
• To assess the efficacy of this new method in improving prediction accuracy for complex traits and G×E compared to existing benchmarks.
• To demonstrate the potential of combining statistical WGP with mechanistic biological models.

Main Methods:

• Developed a novel WGP framework incorporating CGMs via ABC to estimate whole genome marker effects.
• Utilized synthetic datasets to test the performance of the new approach against the standard GBLUP method.
• Evaluated prediction accuracy for traits with non-additive gene effects in both observed and unobserved environments.

Main Results:

• The novel ABC-CGM-WGP approach demonstrated superior accuracy compared to the benchmark GBLUP method.
• Improved predictions were observed for traits influenced by non-additive gene effects.
• The method showed effectiveness in predicting performance across both familiar and novel environments.

Conclusions:

• Integrating biological knowledge, specifically CGMs, into WGP using ABC is a promising novel strategy.
• This approach significantly enhances prediction accuracy for challenging traits and G×E scenarios in plant breeding.
• The proof of concept validates the potential of ABC-CGM-WGP for advancing applied genetics and breeding programs.