Genomic Selection for Pea Grain Yield and Protein Content in Italian Environments for Target and Non-Target Genetic Bases

Enhanced pea cultivation, which can increase the sustainability of European agriculture, requires better-performing cultivars. This study investigated the genomic selection (GS) ability to predict grain yield, protein content, and protein yield on the same or a different genetic base (target/non-target GB) relative to that employed for model training. GS models were developed on 276 lines from three Recombinant Inbred Line (RIL) populations evaluated in three Italian autumn-sown environments using 5537 SNPs from genotyping by sequencing. Validation in two cropping years concerned 108 independent lines from five RIL populations, of which two belonged to the GS training set, and three shared one parent each with training populations. A genome-wide association study performed on the GS training set using 18,674 SNPs highlighted the polygenic control of protein content and grain yield, with several environment-dependent QTLs for yield. Intermediate/high predictive ability within or across populations emerged for all traits in the target GB (0.359-0.675), with some variation depending on the population. Predictive ability in the non-target GB was modest/intermediate for protein content, and null/poor for the other traits. No inverse correlation emerged between grain yield and protein content. GS proved useful for all traits in the target GB and for protein content in a non-target GB.