Combining phenomic and genomic selection for pea breeding improvement

Abstract Pea ( Pisum sativum L.) is a strategic crop in the development of sustainable agriculture. However, the genetic gain remains limited despite advances in breeding. Genomic selection holds promise to accelerate varietal improvement, but its high implementation cost restricts its use in crops. Phenomic selection, based on near‐infrared spectroscopy data, is a cost‐effective alternative demonstrated in various crops, but not yet undertaken in pea. This study aims to assess the predictive ability of phenomic selection, alone and combined with genomic selection, for yield‐related traits in a panel of elite spring pea lines evaluated across 12 environments. Three cross‐validation scenarios were implemented to simulate predictions across different years and locations. Our results show that phenomic selection is as effective as genomic selection at predicting yield. The integrative model, combining spectral and molecular data, consistently achieved the highest accuracy for most traits, particularly for complex traits such as seed yield and seed protein yield. In temporal prediction scenarios, the most accurate predictions were obtained using the spectra data from the same year as phenotyping. In spatial prediction scenarios, predictive accuracy varied by site and year, nevertheless, integrative phenomic‐genomic models consistently outperformed univariate approaches. These findings confirm the potential of phenomic selection in pea and underscore the added value of combining near‐infrared spectroscopy and genotyping data to improve the prediction of complex traits in breeding programs. In the face of increasing environmental variability, the integrative approach offers a valuable tool for accelerating genetic gain.