[关键词]
[摘要]
杂交种鉴定是玉米杂交育种的一个重要环节。全基因组选择(genomic selection,GS)能够利用影响性状的所有变异位点估计育种值并进行有效选择,可以加快优良杂交种的选育进程。本研究利用123份自选系和8个测验种以NCII遗传交配设计组配的537个F1杂交种为材料,结合玉米5.5 K液相育种芯片鉴定的11734个高质量SNP标记,利用5种GS方法的加性模型和加-显模型对新乡、周口环境穗长、穗粗和穗行数以及最佳线性无偏估计(best linear unbiased estimate,BLUE)值开展预测研究,以期为利用GS技术筛选玉米优良杂交种提供指导。结果表明,穗行数的预测准确性最高,为0.76~0.84,穗粗次之,为0.54~0.65,穗长的预测精度最低,为0.33~0.54。3个性状均利用BLUE值估计的准确性最高。在周口环境,相比加性模型,基因组最佳线性无偏预测和3种贝叶斯方法加-显模型对穗长预测准确性的提高率为9.09%,穗粗3种贝叶斯方法加-显模型的提高率为1.85%。绝大多数情况下,加性模型和加-显模型的预测准确性相等,说明整合显性效应并不能提高杂交种穗长、穗粗和穗行数基因组预测精度。综合考虑预测准确性方差分析、多重比较和聚类分析,再生核希尔伯特空间是最佳的杂交种预测方法。
[Key word]
[Abstract]
Hybrid identification is an important step in maize hybrid breeding. Genomic selection (GS) can be used to estimate breeding values and make effective selection using all variation loci affecting traits, which can accelerate the selection process of excellent hybrids. In this study, 123 self-developed lines and eight tester lines were used to produce 537 F1 hybrids according to the NCII genetic mating design. Combined with 11734 high-quality SNP markers identified by maize 5.5 K liquid breeding chip, the additive and additive-dominance model of five GS methods were used to predict ear length, ear diameter, and kernel row number of Xinxiang, Zhoukou, and best linear unbiased estimate (BLUE) value, in order to provide a guidance for selecting maize excellent hybrids using GS technique. The results showed that the prediction accuracy of kernel row number was the highest (0.76-0.84), followed by ear diameter (0.54-0.65), and that of ear length was the lowest (0.33-0.54). The accuracy of genomic breeding value estimated by BLUE values was the highest for all three traits. Compared with the additive model, genomic best linear unbiased prediction and three Bayesian methods with additive-dominance model improved the prediction accuracy of ear length by 9.09%, and the three Bayesian methods improved the prediction accuracy of ear diameter by 1.85% in Zhoukou environment. In most cases, the prediction accuracy of additive and additive-dominance model was equal, indicating that the incorporation of dominant effect could not improve the genomic prediction accuracy of ear length, ear diameter, and kernel row number. Taken the analysis of variance, multiple comparison, and cluster analysis for the prediction accuracy together, reproducing kernel Hilbert space was the best method for hybrid prediction.
[中图分类号]
[基金项目]
河南省科技攻关项目(222102110043; 212102110295)和河南省农业科学院优秀青年基金(2020YQ04)