Pipelines for selecting conditional cores and iterative genomic prediction for manuscript: Development of Conditional Cores Via Super Saturated Designs for Genetic Diversity-Based Prediction Models

Genetic diversity is fundamental to sustainable crop improvement. This study uses 1,755 recombinant inbred lines from the SoyNAM population that were phenotyped for yield across nine environments to explore how training core composition based on genetic diversity affects the predictive ability of genomic selection models. Using supersaturated designs (SSDs), we quantified population diversity and generated training populations that were optimized conditionally for fixed test sets. We deployed two alternative strategies, one that maximizes and one that minimizes genetic diversity between training and test populations. Our results show that maximizing genetic diversity improves prediction accuracy compared to a random baseline in simple GBLUP models. However, this advantage disappears when explicit family pedigree information is available. We also found that maximizing genetic diversity preserves at least one individual from as many different families as possible, while minimizing diversity eliminates or overrepresentes them.