High-throughput phenotyping is boosting plant breeding by generating large-scale phenotypic data for traits that were previously expensive and/or time-consuming to measure. A high-throughput phenotyping platform integrating hyperspectral imaging with a Python workflow has been developed to phenotype nutritional components in almond breeding populations, addressing the current phenotyping bottleneck of conventional methods. Kernel and powder samples from a reference set of 112 almond genotypes were scanned using a hyperspectral camera in the SWIR range (900–1700 nm) and subsequently analysed for nutritional components, including fats, protein, fiber, sucrose, fatty acids, and phytosterols. Partial Least Squares (PLS) models were developed to predict nutritional components in almond kernels, achieving cross-validation RMSE (RMSE_CV) values of 0.73, 1.28, 7.90, and 1.96, and corresponding R_CV² values of 0.82, 0.86, 0.66, and 0.57 for protein, fats, β-sitosterol, and oleic acid (C18:1), respectively. Selected PLS models were implemented to predict the nutritional components of 528 genotypes from a germplasm collection and six F₁ populations. Narrow-sense heritability for these predicted traits was estimated using an advanced linear mixed model incorporating pedigree and genomic data using the 60K Almond SNP array, revealing relevant additive effects for predicted traits (h²>0.5). The approach employed here represents a major advance in nutritional almond breeding, enabling the phenotyping of six times more individuals than previous studies and generating the largest phenotypic dataset of nutritional components in almonds and other tree nuts.

Almond [Prunus dulcis (Mill.) D.A. Webb] is one of the oldest cultivated nut tree crops in the world. Breeding efforts aim to develop new cultivars with improved economically significant traits, including higher yields and superior kernel quality, which have a complex genetic architecture. Almond yield has been linked to traits such as flowering or fruit set, among others. Despite the importance of flowering time, its genetic control is still not fully understood in almond trees. Recent advancements in genomics, including the new almond Axiom 60K SNP array and the availability of transcriptomic resources have enabled researchers to integrate various findings from QTL analyses, genome-wide association studies (GWAS) and gene expression and protein network interactions. This study examined 194 individuals, including 80 genotypes from the CEBAS-CSIC germplasm collection and an F₁ progeny with 114 descendants from the cross between the French selection ‘R1000’ and the Spanish cultivar ‘Desmayo Largueta’. We obtained an ultra-high-density genetic map with 4583 SNPs, with a density of 0.46 cM/marker and a total length of 509.29 cM. We identified 29 QTLs: 10 for flowering time (FT) and 19 for blooming density (BD). The GWAS identified significant associations for 23 SNP traits using different models. A total of 35 SNPs exhibited identical allelic effect patterns in the F₁ and the germplasm. Using SNPeff, we detected 104 distinct effects. Combined with re-analyzed RNA-seq data and protein–protein interaction analyses, the results allowed us to detect a list of candidate genes, including Prudul26A008435 (L-ascorbate oxidase-like), Prudul26A008619 (UDP-glucose 4-epimerase), and Prudul26A007993 (gibberellin 2-β-dioxygenase), which are associated with flowering time as well as Prudul26A003733 (Aminotransferase class Ⅳ), which is linked to blooming density. This information will contribute to a comprehensive understanding of flowering, a crucial almond trait.