Fig. 1: Using correlation between additive and non-additive p-values for efficient non-additive GWAS.

Hit in the legend is defined as a variant that is genome-wide significant in a non-additive model but does not reach this significance threshold in an additive model. Non-hits are not genome-wide significant in any model. A two-proportion Z-test was used to calculate p-values. A–B Examples of simulation of additive and non-additive p-values for a variant with known true recessive (A) or dominant (B) effects and MAF = 0.3. While additive p-values do not reach the genome-wide significance level, they are still relatively small. C–D Simulated additive and non-additive p-values produced for variants with MAF = 0.3 and with variable recessive (C) and dominant (D) true effects. Every simulated variant on the scatter plot is not genome-wide significant in the additive model (\({p}_{{add}} > 5\times {10}^{-8}\)). Blue points correspond to variants reaching genome-wide significance in the non-additive model, while red points show variants that are not genome-wide significant in non-additive models. The density above the scatterplot shows the distribution of additive p-values for variants that reach genome-wide significance exclusively in a non-additive model and can be used to determine the correspondence between the discovery rate and the choice of additive p-value threshold. Bright blue and bright red represent variants retained for non-additive analysis after pre-filtering using the additive p-value threshold. Transparent blue and transparent red indicate variants that were filtered out.