摘要
随着小麦等粮食供需矛盾的日益突出,提高作物产量刻不容缓,在影响小麦产量的基因筛选育种领域仍有很多问题亟待研究。该文利用全基因组选择(Genomic Selection,GS)研究冬小麦的基因型和表型数据之间的量化关系,将5种机器学习模型(Linear-SVR、RBF-SVR、Ridge、LightGBM、XGBoost)与2种传统育种模型(GBLUP、BayesA)进行对比,对新育种群体进行表型(如产量、株高、千粒重)预测和选择,进而找到高效的人工智能(AI)算法用于筛选出影响小麦性状的关键基因。研究结果显示,GBLUP,Ridge,Linear-SVR对于小麦产量具有较高的预测准确性,因此机器学习模型结合传统GBLUP模型能够提高基因型预测的准确性,从而为人工智能算法应用于小麦全基因组选择育种开辟新的道路、提供有益的技术支持。
With the contradiction between supply and demand of wheat and other foods becoming increasingly prominent,it is urgent to improve crop yield.There are still many problems to be studied in the field of gene selection and breeding that affect wheat yield.In this paper,whole genome selection(GS)was used to study the quantitative relationship between genotypic and phenotypic data of winter wheat.Five machine learning models(i.e.Linear-SVR,RBF-SVR,Ridge,LightGBM,and XGBoost)were compared with two traditional breeding models(i.e.GBLUP and BayesA)to predict and select phenotypes(such as yield,plant height,weight per 1000 seeds)of new breeding populations.Then an efficient artificial intelligence(AI)algorithm was found to screen out the key genes affecting wheat traits.The results show that GBLUP,Ridge and Linear-SVR have high prediction accuracy for wheat yield,so the machine learning model combined with traditional GBLUP model can improve the accuracy of genotype prediction,so as to open up a new way and provide useful technical support for the application of artificial intelligence algorithm in wheat genome selection and breeding.
出处
《智慧农业导刊》
2022年第19期4-6,共3页
JOURNAL OF SMART AGRICULTURE
基金
省级大学生创新创业训练计划项目(S202110451258)
鲁东大学专创融合课程建设重点项目(202114)。
关键词
全基因组选择
人工智能算法
机器学习
分子育种
冬小麦
whole genome selection
artificial intelligence algorithm
machine learning
molecular breeding
winter wheat
