Genomic prediction and association mapping of early season flood tolerance in soybean

Flooding has become a major threat to soybean [Glycine max (L.) Merr.] production as the frequency and intensity of extreme precipitations have been increasing due to climate change. While advances have been made in identifying soybean genetic resources and genomic regions associated with mid-season flood tolerance, there is limited understanding of early season flood tolerance at the vegetative growth stages V2/V4. This study aimed to identify genomic regions associated with early season flood tolerance using a diverse panel of 254 soybean accessions, as well as investigate the viability of implementing genomic prediction models for flood tolerance. Field trials were conducted over 2 years, with flooding imposed at the V2/V4 vegetative growth stages. Genome-wide association studies were performed using the Bayesian-information and linkage-disequilibrium iteratively nested keyway and the multiple locus mixed linear model. Forward stepwise genomic prediction models using random forest (RF) were developed to identify the set of single nucleotide polymorphisms (SNPs) yielding the highest prediction accuracy while assessing the negative impacts of multicollinearity and overfitting on prediction accuracy. Genomic regions on chromosomes 4, 17, and 20 associated with early season flood tolerance were identified, all distinct from regions previously identified for mid-season tolerance. The RF model achieved a prediction accuracy of 0.64 with 29 selected SNPs, significantly improving over RF and ridge regression best linear unbiased prediction models with higher SNP counts. These findings provide genomic tools for improving the efficiency of breeding for early season flood tolerance, supporting the need to develop season-long flood-tolerant soybean genotypes.