Genomic selection predicts unknown phenotypes by using high-density genetic markers covering the genome.In the plant,this method allows early selection for traits,retaining dominant individuals and saving costs for field management and phenotype identification,which greatly accelerating the breeding process.In this study,genomic,transcriptomic,and metabolomic data were used for genomic prediction of agronomic and quality traits of maize by using two statistical models,rrBLUP,and LASSO.We found that the order of predictive power was genomic data,transcriptomic data,and metabolomic data.For different traits,genomic prediction was more powerful than agronomic traits for quality traits.For both rrBLUP and LASSO models,rrBLUP was the best model for all traits when using genomic data,53 traits were the best predicted by rrBLUP and 2 traits were the best predicted by LASSO when using transcriptomic data,43 traits were the best predicted by rrBLUP and 12 traits were the best predicted by LASSO,and 12 traits were the best predicted by LASSO based on metabolomic data.In addition,when per-forming genomic prediction using different lineages,the accuracy of predicting the temperate maize from the tropic maize was slightly better than that of predicting the tropic maize from the temperate.For quality traits,we found the cross-lineage prediction was higher than the within-lineage prediction.This study systematically evaluated the differences in the predictive ability of maize agronomic and quality traits based on various multi-omics data and statistical models,which providing a theoretical basis for future genomic breeding of important agricultural traits in maize.
关键词
玉米/农艺和品质性状/基因组预测/多维组学数据
Key words
maize/agronomic and quality trait/genomic prediction/multi-omics data
维普
万方数据