Many studies have investigated the association between DNA methylation alterations and disease occurrences using two design paradigms, traditional case-control and disease-discordant twins. In the disease-discordant twin design, the affected twin serves as the case and the unaffected twin serves as the control. Theoretically the twin design takes advantage of controlling for the shared genetic make-up, but it is still highly debatable if and how much researchers may benefit from such a design over the traditional case-control design.

In this study, we investigate and compare the power of both designs with simulations. A liability threshold model was used assuming that identical twins share the same genetic contribution with respect to the liability of complex human diseases. Varying ranges of parameters have been used to ensure that the simulation is close to real-world scenarios. Our results reveal that the disease-discordant twin design implies greater statistical power over the traditional case-control design. For diseases with moderate and high heritability (>0.3), the disease-discordant twin design allows for large sample size reductions compared to the ordinary case-control design. Our simulation results indicate that the discordant twin design is indeed a powerful tool for epigenetic association studies.

Computer scripts are available at https://github.com/zickyls/EWAS-Twin-Simulation.

Supplementary data are available at Bioinformatics online.