结合辅助性状的玉米全基因组选择预测力评估

Predictability of maize genome-wide selection combined with auxiliary traits

多性状联合全基因组选择能够有效利用性状间的遗传相关和环境相关,有望提高表型预测的准确性。本研究提出了结合辅助性状的全基因组选择策略,以来源广泛的342份玉米自交系为试验材料,对其进行基因分型测序(GBS)并分析其农艺性状,对每个目标性状均基于辅助性状及其组合进行预测,利用五倍交叉验证法评价其预测力。结果表明,利用与目标性状相关性较高的辅助性状可较大程度地提升预测力,尤其是对于低遗传力性状;随着辅助性状个数的增加,预测力也随之增加。进一步比较了5种统计模型结合辅助性状的全基因组选择的表型预测力,总体而言,再生核希尔伯特空间(RKHS)模型和贝叶斯B(BayesB)模型的预测效果较优,而极端梯度提升(XGBOOST)模型的预测效果较差。本研究结合辅助性状有效提高了玉米全基因组选择的预测准确性,为玉米的全基因组选择育种提供新的思路和参考。

Multi-trait genomic selection can use genetic and environmental correlations between traits,which holds great promise to improve the prediction accuracy.This study proposed a genomic prediction strategy using auxiliary traits.A total of 342 maize inbred lines from a diversity panel were used as test materials.Genotyping by sequencing(GBS)was performed and six agronomic traits were measured in the field.Each target trait was predicted based on auxiliary traits and their combinations.The predictability was evaluated using five-fold cross-validation.The results showed that the use of auxiliary traits highly correlated with target traits greatly improved predictability and low-heritability traits could benefit more from auxiliary traits.As the number of auxiliary traits increased,the predictability also increased.We also compared the prediction performance of five different models combined with auxiliary traits.Overall,reproducing kernel Hilbert space(RKHS)model and BayesB model performed well,while extreme gradient boosting(XGBOOST)model performed worst.This study improves the accuracy of genomic prediction and provides new ideas and references for genomic selection breeding of maize.