Genomic Selection Paves Way for the Identification of Rust Disease Resistant Genotypes in Bread Wheat (Triticum aestivum).

In the last two decades, genomic prediction (GP) or Genomic Selection (GS) methods have been widely adopted in various plant and animal breeding programs globally. GP/GS is a promising method that employs genomic markers to calculate genomic-estimated breeding values (GEBVs) to select best individuals. To evaluate the performance of different genomic selection (GS) models, we examined six different models namely, ridge regression (RR), least absolute shrinkage and selection operator (LASSO), genomic best linear unbiased prediction (GBLUP), elastic net (EN), reproducing kernel Hilbert spacing (RKHS), and random forest (RF) models, for seedling and adult plant resistance to leaf, stem and stripe rust of wheat using a panel of 347 wheat germplasm accessions. The GBLUP and RF models performed noticeably better than the other GS models, with mean predictive abilities of 0.5 and 0.4 for seedling resistance and 0.4 and 0.3 for adult plant resistance (APR) for leaf and stem rust, respectively. Unfortunately, except for a few environments, the performance of GP models in the current study is quite low for stripe rust for both seedling and APR. The outcomes of this study revealed the capability of GP to be applied for breeding initiatives aimed at developing wheat varieties resistant to rust diseases. Moreover, based on favorable allele analysis we also identified a total of 2 lines (CRP-165/42, HGP1-470) that showed resistance to most of the pathotypes at seedling and adult plant stage to all three rusts. These lines can serve as valuable resources for future breeding programs focused on rust resistance.

Keywords: GS; GEBVs; leaf rust; stem rust; stripe rust; seedling resistance; APR

近二十年来，基因组预测（Genomic Prediction, GP）或基因组选择（Genomic Selection, GS）方法已在全球各类动植物育种项目中得到广泛应用。GP/GS是一种极具应用前景的技术，通过利用基因组标记计算基因组估计育种值（Genomic Estimated Breeding Values, GEBVs）以筛选最优个体。为评估不同基因组选择（GS）模型的性能，本研究基于347份小麦种质资源材料构成的试验群体，针对小麦对叶锈病、茎锈病及条锈病的苗期和成株抗性，测试了6种不同模型，分别为岭回归（Ridge Regression, RR）、最小绝对收缩与选择算子（Least Absolute Shrinkage and Selection Operator, LASSO）、基因组最佳线性无偏预测（Genomic Best Linear Unbiased Prediction, GBLUP）、弹性网络（Elastic Net, EN）、再生核希尔伯特空间（Reproducing Kernel Hilbert Spacing, RKHS）及随机森林（Random Forest, RF）模型。结果显示，GBLUP与RF模型的表现显著优于其余GS模型：针对叶锈病与茎锈病，二者的苗期抗性平均预测能力分别为0.5与0.4；成株抗性（Adult Plant Resistance, APR）的平均预测能力则分别为0.4与0.3。遗憾的是，本研究中针对条锈病的苗期抗性与成株抗性，GP模型的表现均较差，仅少数环境例外。本研究结果证实，GP技术可应用于培育抗锈病小麦品种的育种计划。此外，通过优势等位基因分析，本研究共筛选出2份小麦材料（CRP-165/42、HGP1-470），它们在苗期与成株阶段对三种锈病的多数致病型均表现出抗性。这些材料可作为未来锈病抗性育种项目的宝贵种质资源。

关键词：基因组选择（GS）；基因组估计育种值（GEBVs）；叶锈病；茎锈病；条锈病；苗期抗性；成株抗性（APR）