Abstract
Genotyping cost is one of the major limiting factors for wide adoption of genomic selection in wheat. The present study compared two genotyping platforms, a low-density 4K single nucleotide polymorphism (SNP) chip and a Wheat iSelect 90K SNP, in genomic prediction of grain yield (YLD) and yield related traits in a training population composed of 224 spring wheat lines, and a new population composed of 141 spring wheat lines, with and without consideration of a fixed effect for major plant adaptation genes-derived competitive allele specific PCR (KASP) markers. Five models were evaluated and compared, which included 4K SNP-only, 4K SNP + KASP, 90K SNP-only, 90K SNP + KASP, and KASP-only. Within the training population, the 90K array showed slightly higher predictive ability for YLD, whereas the 4K panel performed better for thousand kernel weight (TKW). Incorporating KASP fixed effects, the 4K SNP + KASP and 90K SNP + KASP models significantly improved prediction for YLD, heading date (HD), and plant height (PHT) compared to the corresponding SNP-only models. KASP-only outperformed both SNP-only models for HD, but for PHT it only outperformed the 4K SNP-only model. In cross-population prediction, KASP-only models outperformed 4K SNP-only for HD and PHT. These results suggest that breeding programs can deploy genomic selection with a mid-density marker platform, especially, the 4K SNP + KASP model offers a cost-effective strategy for genomic prediction, while KASP-only genotyping provides a practical ultra-low-cost option for screening heading date and plant height.