Targeted expansion of a barley genebank core collection facilitates the discovery of disease resistance loci.

Abstract

Utilizing the diversity preserved in genebank collections is essential for accelerating crop improvement, yet information is often limited to selected core collections. Genome-wide prediction (GWP) offers a promising approach to large-scale phenotypic imputation, with proven utility in practical pre-breeding contexts. In this study, we leveraged GWP to expand the German Federal ex situ barley core collection (core1000) with a focus on resistance to Puccinia hordei, Blumeria graminis hordei, and Rhynchosporium commune. Using the barley core1000 collection, which was originally selected to maximize molecular diversity, we trained genomic prediction models and imputed resistance scores for 20,458 genebank accessions based on sequence data encompassing 306,049 high-quality SNPs. To empirically validate prediction accuracy, we selected 300 spring and winter barley genotypes for field evaluation across four environments, resulting in moderate-to-strong correlations between predicted and observed resistance levels. Genome-wide association mapping in this set revealed five marker-trait associations that were not detected in the original core1000 collection. These results demonstrate that prediction-informed sampling can effectively expand trait-relevant genetic diversity and increase the frequency of resistance-associated alleles, thereby improving the power to detect loci that may be overlooked in conventional panels. Accordingly, GWP supports the targeted inclusion of accessions with trait-relevant variation and enhances the value of genebank resources for trait discovery and pre-breeding applications.

Plant phenotyping relevance

ゲノム情報から病害抵抗性という植物形質を大規模に推定・補完する方法を開発し、圃場評価で予測精度を検証しているため、方法論が中心である。

Code and data availability

The authors deposited the paper's raw disease-resistance phenotypic data, BLUEs for the spring/winter validation set, and the R script for curation/heritability/BLUE computation in the public e!DAL-PGP repository (Yuan 2025). The companion IPK/2024/7 dataset belongs to cited prior work (Yuan et al. 2025 training data).