Abstract
DNA methylation plays an important role in normal human development and disease. In epigenome-wide association studies (EWAS), a univariate test for association between a phenotype and each cytosine-phosphate-guanine (CpG) site has been widely used. Given the number of CpG sites tested in EWAS, a stringent significance cutoff is required to adjust for multiple testing; in addition, multiple nearby CpG sites may be associated with the phenotype, which is ignored by a univariate test. These two factors may contribute to the power loss of a univariate test. As an alternative, we propose applying an adaptive gene-based test that is powerful in genome-wide association studies (GWAS), called aSPUw, to EWAS for simultaneous testing on multiple CpG sites within or near a gene. We show its application to the GAW20 methylation data set.
Highlights
DNA methylation of cytosine residues at cytosine-phosphate-guanine (CpG) dinucleotides is of particular interest because it has a central role in the normal human development and disease [1]
Given the number of CpG sites tested in Epigenome-wide association studies (EWAS), a univariate test must meet a stringent threshold for statistical significance; in addition, a univariate test does not take advantage of possible existence of multiple associated CpG sites within a gene
We tested the association between the log pretreatment fasting TGs and the methylation levels of the CpG sites within each gene’s coding region; we used the identity function as the link function in generalized linear mixed model (GLMM)
Summary
DNA methylation of cytosine residues at cytosine-phosphate-guanine (CpG) dinucleotides is of particular interest because it has a central role in the normal human development and disease [1]. Epigenome-wide association studies (EWAS), analogous to genome-wide association studies (GWAS), are becoming increasingly popular to interrogate methylation changes associated with a disease or related environmental factors [2]. The common statistical analysis in EWAS uses a single marker test for association between a phenotype and each of the CpG sites. Given the number of CpG sites tested in EWAS, a univariate test must meet a stringent threshold for statistical significance (for example, p value < 1 × 10− 7 often used for the Illumina 450 K array); in addition, a univariate test does not take advantage of possible existence of multiple associated CpG sites within a gene. We consider statistical methods that test for the association between multiple CpG sites in a gene and a phenotype simultaneously
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
