Table 3.

Comparison of Enrichment Estimates by EM-DAP1 and EM-MCMC after a Single Iteration in Analysis of GEUVADIS Data

Method	Footprint SNPs		Binding Variants
α	95% C.I.	α	95% C.I.
EM-MCMC	0.14	(0.04, 0.24)	0.39	(0.32, 0.49)
EM-DAP1	0.12	(−0.01, 0.25)	0.41	(0.30, 0.51)

The binding SNPs refer to the genetic variants that are computationally predicted to disrupt TF binding, and the footprint SNPs are those simply located in the DNaseI footprint region but not predicted to affect TF binding. The enrichment estimates from both methods are very similar. The MCMC algorithm accounts for multiple independent association signals and yields slightly tighter confidence intervals, as expected. However, the EM-DAP1 is much more computationally efficient: it runs almost one thousand times faster than the EM-MCMC algorithm.