Müller Hans-Helge, Pahl Roman, Schäfer Helmut
Institute of Medical Biometry and Epidemiology, University of Marburg, Marburg, Germany.
Genet Epidemiol. 2007 Dec;31(8):844-52. doi: 10.1002/gepi.20245.
We propose optimized two-stage designs for genome-wide case-control association studies, using a hypothesis testing paradigm. To save genotyping costs, the complete marker set is genotyped in a sub-sample only (stage I). On stage II, the most promising markers are then genotyped in the remaining sub-sample. In recent publications, two-stage designs were proposed which minimize the overall genotyping costs. To achieve full design optimization, we additionally include sampling costs into both the cost function and the design optimization. The resulting optimal designs differ markedly from those optimized for genotyping costs only (partially optimized designs), and achieve considerable further cost reductions. Compared with partially optimized designs, fully optimized two-stage designs have higher first-stage sample proportion. Furthermore, the increment of the sample size over the one-stage design, which is necessary in two-stage designs in order to compensate for the loss of power due to partial genotyping, is less pronounced for fully optimized two-stage designs. In addition, we address the scenario where the investigator is interested to gain as much information as possible, however is restricted in terms of a budget. In that we develop two-stage designs that maximize the power under a certain cost constraint.
我们提出了用于全基因组病例对照关联研究的优化两阶段设计，采用假设检验范式。为节省基因分型成本，仅在一个子样本中对完整的标记集进行基因分型（第一阶段）。在第二阶段，然后在其余子样本中对最有前景的标记进行基因分型。在最近的出版物中，提出了两阶段设计，可将总体基因分型成本降至最低。为实现完全的设计优化，我们还将抽样成本纳入成本函数和设计优化中。所得的最优设计与仅针对基因分型成本进行优化的设计（部分优化设计）显著不同，并实现了进一步的大幅成本降低。与部分优化设计相比，完全优化的两阶段设计具有更高的第一阶段样本比例。此外，对于完全优化的两阶段设计，为弥补因部分基因分型导致的效能损失而在两阶段设计中必需的样本量相对于单阶段设计的增量并不那么明显。此外，我们还探讨了研究者希望获取尽可能多的信息，但受预算限制的情况。在此，我们开发了在特定成本约束下使效能最大化的两阶段设计。
