Combining genome-wide prediction and a phenology model to simulate heading date in spring barley

详细信息    查看全文

• 作者：Ralf Uptmoor^a ; ^{ralf.uptmoor@uni-rostock.de} ; Klaus Pillen^b ; Claudia Matschegewski^a
• 关键词：Photoperiod ; Temperature ; Heading date ; Genomic prediction ; Phenology model
• 刊名：Field Crops Research
• 出版年：2017
• 出版时间：15 February 2017
• 年：2017
• 卷：202
• 期：Complete
• 页码：84-93
• 全文大小：1746 K
• 卷排序：202

文摘

Phenotype by genotype prediction based on ecophysiological models, which account for allelic gene, QTL, or marker effects, have many possible applications in plant breeding programs. The goal of the present study was to predict heading date of individual lines of a Hordeum vulgare x H. vulgare ssp. spontaneum BC2DH-population using a phenology model parameterized with marker effects derived from ridge regression best linear unbiased prediction. The genetic linkage map included SSR markers and flowering-time genes. Effects of photoperiod and temperature on heading date were measured under controlled conditions on a subset of the population comprising the recurrent parent and 36 BC2DH candidate introgression lines covering the H. spontaneum genome. Marker effects, which were subsequently used for model parameterization, were estimated. Model evaluation was carried out on already published field trial data comprising the 36 BC2DH lines and 266 independent BC2DH lines from the same cross. Applying the model on the lines used for model parameterization explained 33–51% of heading-date variation in three of the four evaluation environments but only 20% of the variation in the fourth environment. Heading dates of the 266 independent lines were predicted with less accuracy. Between 20 and 25% of phenotypic variation was explained by the model in three environments and only 8% of heading date variation in the fourth environment. The root mean squared error (RMSE) was slightly higher for independent lines than for the lines used for model parameterization. Dissecting RMSE into its components revealed that RMSE was largely influenced by a systematic bias in most environments and by the missing ability of the model to describe the observed variation within the set of genotypes in all environments. Comparing the combined genome-wide prediction (GWP) and phenology model with a conventional GWP model gave similar prediction accuracies if the training set had the same size.