Background: Pea (Pisum sativum) is a prevalent cool-season crop that produces seeds valued for their high protein content. Modern cultivars have incorporated several traits that improved harvested yield. However, progress toward improving seed quality has received less emphasis, in part due to the lack of tools for easily and rapidly measuring seed traits. In this study we evaluated the accuracy of single-seed near-infrared spectroscopy (NIRS) for measuring pea-seed weight, protein, and oil content. A total of 96 diverse pea accessions were analyzed using both single-seed NIRS and wet chemistry methods. To demonstrate field relevance, the single-seed NIRS protein prediction model was used to determine the impact of seed treatments and foliar fungicides on the protein content of harvested dry peas in a field trial.
Results: External validation of partial least squares (PLS) regression models showed high prediction accuracy for protein and weight (R2 = 0.94 for both) and less accuracy for oil (R2 = 0.74). Single-seed weight was weakly correlated with protein and oil content in contrast with previous reports. In the field study, the single-seed NIRS predicted protein values were within 10 mg g-1 of an independent analytical reference measurement and were sufficiently precise to detect small treatment effects.
Conclusion: The high accuracy of protein and weight estimation show that single-seed NIRS could be used in the dual selection of high-protein, high-weight peas early in the breeding cycle, allowing for faster genetic advancement toward improved pea nutritional quality. © 2020 Society of Chemical Industry.
Keywords: Pisum sativum; high-throughput phenotyping; near-infrared spectroscopy; nutritional quality; seed protein; single-seed phenotyping.
© 2020 Society of Chemical Industry.