The most popular beverage worldwide, coffee, is responsible for a billionaire market chain with arabica coffee leading the production. Coffee breeding programs are focusing on high yield, excellent beverage quality, and disease resistance, but the bienniality comes to a challenge to overcome bean production. The bienniality, the seasonal variation between high and low yielding, is a genetically controlled physiological event that affects yield stability in arabica coffee. However, there are no studies on the best strategies to implement genomic selection in coffee, including how to establish training populations and deal with the biennially. Thus, the objective was evaluated the potential of genomic selection applied to arabica coffee, with particular consideration on how to estimate bienniality effect on genomic prediction accuracy for yield. The population (n = 586) high-density genotyped by GBS was measured in the low (2005 and 2007), and high (2006 and 2008) yield years. The genomic prediction models were established considering genotype and genotype × year effects. Different prediction scenarios were proposed, considering single-year training sets and grouping the data according to bienniality. Overall, training genomic models on biennium of successive years, and predicting the following biennium appears to be the most effective strategy between all tested scenarios. The comparison of phenotypic and prediction approaches revealed an increased selection response using genomic selection, mainly due to the reduced time per breeding cycle. These results can shed light on the implementation of a genome-based selection of arabica coffee and lead to more efficient breeding strategies.

中文翻译：

---

两年期对阿拉比卡咖啡产量基因组预测的影响

世界上最受欢迎的饮料咖啡负责一个亿万富翁的市场链，其中阿拉比卡咖啡是主导生产的。咖啡育种计划的重点是高产、优质的饮料质量和抗病性，但两年期对克服咖啡豆生产提出了挑战。两年期，即高产和低产之间的季节性变化，是一种遗传控制的生理事件，会影响阿拉比卡咖啡的产量稳定性。然而，没有关于在咖啡中实施基因组选择的最佳策略的研究，包括如何建立训练种群和每两年处理一次。因此，目标是评估应用于阿拉比卡咖啡的基因组选择的潜力，特别考虑如何估计两年期对产量基因组预测准确性的影响。在低（2005 年和 2007 年）和高（2006 年和 2008 年）产量年份测量了由 GBS 进行基因分型的高密度种群（n = 586）。考虑基因型和基因型×年效应建立基因组预测模型。考虑了单年训练集并根据两年期对数据进行分组，提出了不同的预测方案。总体而言，连续几年训练基因组模型并预测下一个两年期似乎是所有测试场景中最有效的策略。表型和预测方法的比较揭示了使用基因组选择的选择反应增加，主要是由于每个育种周期的时间减少。这些结果可以阐明基于基因组的阿拉比卡咖啡选择的实施，并导致更有效的育种策略。