Fig. 4
Disease odds ratios conditioned on the genotype of 2 causal loci: Odds ratio effect sizes conditioned on pure epistatic pairs of loci for disease status in the simulated case-control datasets. Causal effects for DSL 1 and DSL 2 A are conditioned on allele A for DSL 1. The risk allele A of DSL 2 A only increases risk for individuals carrying at least one copy of the DSL 1 risk allele (DSL 1 is epistatic to DSL 2 A). The low risk CC/TT genotype is set as the baseline (OR=1). The other genotype combinations are coded according to g1,g2 and their product g1×g2. Odds ratio are obtained by exponentiating the β3 coefficient of the interaction term from the logistic regression (see text). Error bars: 95% confidence intervals of possible odds ratio that are obtained in different simulated case-control samples