mmGEBLUP: an advanced genomic prediction scheme for genetic improvement of complex traits in crops through integrative analysis of major genes, polygenes, and genotype-environment interactions.

Abstract

Current genomic prediction (GP) models often fall short of fully capturing the genetic architecture of complex traits and providing practical breeding guidance, particularly under varying environments. Here, we propose the mmGEBLUP, an advanced GP scheme designed to tackle the current limitations in fully exploiting the genetic architecture of complex traits and to predict individual breeding value (BV) with multi-environment trial data. Our approach considers four genetic structural indicators to capture the genetic architectures stepwise across four models: the Genomic Best Linear Unbiased Prediction (GBLUP) model considers only main polygenic effects; the GEBLUP model includes both main and genotype-by-environment (GE) interaction polygenic effects; and the mmGBLUP and mmGEBLUP models further incorporate main and GE interaction effects of major genes. Through systematic simulations and applications to nine traits, three in rice and six in tobacco, we show stepwise increases in prediction accuracy from GBLUP to mmGEBLUP, providing evidence on the scale of heritability and polygenicity of traits. In practical terms, we predict four components of BV: major additive, minor additive, major interaction, and minor interaction. Interestingly, we discover that for traits like natural leaf number in tobacco, the major additive BVs for the top 20 individuals are substantially equal; it is the minor additive BV that causes the difference in the total BV. The relative size of major/minor additive BVs suggests performing either marker-assisted selection or genomic selection or both. Overall, mmGEBLUP is an advanced prediction scheme that enhances the understanding of genetic architectures and facilitate the genetic improvement of complex traits in crops under diverse environments.