Figure 6. Exploring differential effects in 120 directly typed polymorphisms across 70 candidate malaria-protecting genes.

(A) Case-control effect sizes estimated for the ‘severe malaria’ sub-population versus the ‘not severe malaria’ sub-population (n = 3940 controls and n = 2220 cases, with approximately 1279 in the ‘severe malaria’ sub-population and 941 in the ‘not severe malaria’ sub-population). The vertical and horizontal grey lines show the 95% credible intervals. (B) The log₁₀ p-values testing the hypothesis that the effects are the same for the two sub-populations relative to controls. The top dashed line shows the Bonferroni corrected $\alpha =0.05$ significance threshold (assuming 70 independent tests). The bottom dashed line shows the nominal $\alpha =0.05$ significance threshold. In both panels, red circles denote $p<0.05$ (nominal significance level), and red squares denote $p<0.05/70$. (C) Analysis of the rs1050828 SNP (encoding G6PD + 202T) under a non-additive model (hemi/homozygotes and heterozygotes are distinct categories). This shows that heterozygotes are clearly under-represented in the ‘severe malaria’ sub-population and hemi/homozygotes are clearly over-represented in the ‘not severe malaria’ sub-population. (D) Evidence of differential effects for the O blood group (rs8176719, recessive model) and FREM3 (additive model).