Statistica Sinica 19 (2009), 1769-1786

TWO-STAGE GENOME-WIDE ASSOCIATION

STUDIES WITH DNA POOLING

AND GENETIC MODEL SELECTION

Min Yuan$^1$, Yaning Yang$^1$ and Gang Zheng$^2$

$^1$University of Science and Technology of China

and $^2$National Heart, Lung and Blood Institute

Abstract: The two-stage design is a common cost-effective approach for genome-wide association studies. The first stage serves as a screening to identify a subset of single-nucleotide polymorphisms (SNPs) from 100,000 to 500,000 SNPs using a fraction of case-control samples. In the second stage, only the selected SNPs are genotyped using the remaining case-control samples. On the other hand, DNA pooling is another common strategy to save genotyping cost. In this article, we propose a method using DNA pooling in the first stage and genotype-based analysis in the second stage. A joint analysis to combine both stages is applied to a two-stage design with DNA pooling when the underlying genetic model is known. When the genetic model is unknown, we use a robust procedure in the joint analysis by applying genetic model selection in the second stage based on the difference of Hardy-Weinberg disequilibrium coefficients between cases and controls. Performance of our method and comparison with other approaches are investigated by simulation studies.

Key words and phrases: Cost-effective design, DNA pooling, genetic model selection, joint analysis, robustness, trend tests, two-stage.