玉米重要农艺性状的基因组预测分析  

Genomic Prediction Analysis for Maize Important Agronomic Traits

作　　者：马娟 朱卫红[1] 丁俊强[2] MA Juan;ZHU Wei-hong;DING Jun-qiang(Institute of Cereal Crops,Henan Academy of Agricultural Sciences,Zhengzhou 450002;College of Agronomy,Henan Agricultural University,Zhengzhou 450002,China)

机构地区：[1]河南省农业科学院粮食作物研究所,郑州450002 [2]河南农业大学农学院,郑州450002

出　　处：《玉米科学》2022年第1期48-52,共5页Journal of Maize Sciences

基　　金：河南省农业科学院优秀青年基金项目(2020YQ04)。

摘　　要：以玉米自交系郑683-1为轮回亲本,ZPH1388、ZPH5和东237为供体亲本构建BC3F5群体,对穗部和株型性状进行基因组预测分析,研究2种交叉验证方式、6种GS方法和显著数量性状位点(QTL)作为固定效应对预测准确性的影响。结果表明,相比5折交叉验证,10折交叉验证可以提高穗粒数、行粒数、株高和穗位高的预测准确性,降低穗行数和叶夹角的预测能力。相比随机效应模型,穗粒数、穗行数、行粒数和叶夹角将1~2个QTL作为固定效应可以提高基因组估计育种值的准确性,将1~5个QTL作为固定效应会降低穗位高的预测能力。对于株高,贝叶斯A和再生核希尔伯特空间将1~5个QTL作为固定效应可以提高预测的准确性,其他4种方法多数情况固定效应的加入会降低株高的预测能力。在穗粒数和株高中,基因组最佳线性无偏预测的准确性最高,略高于4种贝叶斯方法,再生核希尔伯特空间的准确性最低。对于其他4个性状,最优预测方法受不同交叉验证方式和固定效应模型的影响表现不一。In the present study,an BC3 F5 population that was derived from maize inbred line 683-1(recurrent parent),and donor parents ZPH1388,ZPH5,and Dong237 was used to conduct genomic prediction analysis and to study impacts of two cross-validation methods,six GS methods,and significant quantitative trait loci(QTL)as fixed effects on prediction accuracy for ear and plant type traits.This will provide a guidance for applications of genomic prediction for maize ear and plant type traits.Results showed that compared to five-fold cross-validation,ten-fold cross-validation could improve the prediction accuracy of kernel number per ear,kernel number per row,plant height,and ear height,but could decrease the prediction ability of kernel row number and leaf angel.Compared to random effect model,the incorporation of 1-2 QTL as fixed effects could improve the accuracy of genomic estimated breeding value for kernel number per ear,kernel row number,kernel number per row,and leaf angle,whereas the incorporation of 1-5 QTL as fixed effects could decrease the prediction ability for ear height.For plant height,treating1-5 QTL as fixed effect in Bayes A and reproducing kernel Hilbert space(RKHS)could improve the prediction accuracy,whereas for other methods,the incorporation of fixed effects could reduce the prediction ability in most cases.For kernel number per ear and plant height,the accuracy of genomic best linear unbiased prediction was the highest,which was slightly higher than that of four Bayes methods,and that of RKHS was the least.For other four traits,the best prediction methods varied due to the effects of cross-validation and fixed effect models.

关 键 词：玉米 基因组选择 随机效应模型 固定效应模型 

分 类 号：S513.035.3[农业科学—作物学]