In this study, we used genotype × environment interactions (G×E) models for hybrid prediction, where similarity between lines was assessed by pedigree and molecular markers, and similarity between environments was accounted for by environmental covariables. We use five genomic and pedigree models (M1–M5) under four cross‐validation (CV) schemes: prediction of hybrids when the training set (i) includes hybrids of all males and females evaluated only in some environments (T2FM), (ii) excludes all progenies from a randomly selected male (T1M), (iii) includes all progenies from 20% randomly selected females in combination with all males (T1F), and (iv) includes one randomly selected male plus 40% randomly selected females that were crossed with it (T0FM). Models were tested on a total of 1888 wheat (Triticum aestivum L.) hybrids including 18 males and 667 females in three consecutive years. For grain yield, the most complex model (M5) under T2FM had slightly higher prediction accuracy than the less complex model. For T1F, the prediction accuracy of hybrids for grain yield and other traits of the most complete model was 0.50 to 0.55. For T1M, Model M3 exhibited high prediction accuracies for flowering traits (0.71), whereas the more complex model (M5) demonstrated high accuracy for grain yield (0.5). For T0FM, the prediction accuracy for grain yield of Model M5 was 0.61. Including genomic and pedigree gave relatively high prediction accuracy even when both parents were untested. Results show that it is possible to predict unobserved hybrids when modeling genomic general combining ability (GCA) and specific combining ability (SCA) and their interactions with environments.

中文翻译：

---

使用基因组，谱系和环境协变量相互作用模型的杂种小麦预测。

在这项研究中，我们使用基因型×环境相互作用（G×E）模型进行杂交预测，其中谱系和分子标记评估品系之间的相似性，环境协变量解释环境之间的相似性。我们在四个交叉验证（CV）方案下使用五个基因组和谱系模型（M1-M5）：当训练集（i）包括仅在某些环境（T2FM）中评估的所有雄性和雌性的杂种时，对杂种的预测）不包括来自随机选择的雄性（T1M）的所有后代，（iii）包括来自20％随机选择的雌性与所有雄性（T1F）组合的所有后代，并且（iv）包括一位随机选择的雄性加上40％随机选择的雌性，被交叉（T0FM）。在总共1888个小麦（Triticum aestivum）上测试了模型L.）杂种，连续三年有18位男性和667位女性。对于谷物产量，T2FM下最复杂的模型（M5）的预测准确性比不那么复杂的模型略高。对于T1F，最完整模型的杂种对籽粒产量和其他性状的预测准确性为0.50至0.55。对于T1M，模型M3显示出较高的开花性状预测准确度（0.71），而更复杂的模型（M5）显示出较高的籽粒产量精度（0.5）。对于T0FM，模型M5的谷物产量的预测准确性为0.61。即使父母双方都未经测试，将基因组和谱系包括在内也能提供相对较高的预测准确性。