Multi-trait ridge regression BLUP with de novo GWAS improves genomic prediction for haploid induction ability and agronomic traits of haploid inducers in maize

Abstract

Key message Employing multi-trait and de novo GWAS in a ridge regression BLUP model increases the predictive ability of haploid induction rate of haploid inducers in maize. Ridge regression BLUP (rrBLUP) is a widely used model for genomic selection. Different genomic prediction (GP) models have their own niches depending on the genetic architecture of traits and computational complexity. Haploid inducers have unique trait performances, relevant for doubled haploid (DH) technology in maize (Zea mays L.). We herein compared the performance of single-trait (ST) and multi-trait (MT) GP models (rrBLUP, BayesB, Random Forest, and xGBoost) and employed multi-trait and de novo GWAS in the ridge regression BLUP model for four traits of interest (Days to flowering, DTF; haploid induction rate, HIR; plant height, PHT; primary branch length, PBL) of the multifamily DH inducers (DHIs), and next tested the GP models in multi-parent advanced generation inter-cross (MAGIC) DHIs. The average predictive abilities (PA) of different GP methods across traits were 0.44 to 0.65 in multifamily DHIs. ST/MT de novo GWAS rrBLUP methods increased PA of HIR when using five-fold cross-validation. In addition, MT GP models improved PA by 13% on average across traits relative to ST GP models in MAGIC DHIs. These results provide empirical evidence that employing multi-trait and de novo GWAS in rrBLUP model in genomic selection could benefit the genetic improvement of haploid inducers.