Supplementary Figure 5: Genetic heterogeneity of the severe malaria subtypes cerebral malaria and severe malarial anemia within and across African populations for all loci.

Bar plots show the posterior probability on each of nine models of association where effects on the two subtypes are fixed, independent or correlated within a population combined with being fixed, independent or correlated across all the populations, as indicated by the different colors; the remainder of the posterior probability is on the null model where there is no effect on any of the subtypes. Before seeing the data, the null model is assumed to have 80% probability; the remaining 20% is divided equally among the other models. Approximate Bayes factors (ABFs), calculated as the ratio of the marginal likelihoods of a given model and the null model, are used to compare the evidence between models. (See the Online Methods for details and the specification of priors.) The dashed line at posterior probability = 20% indicates where the combined posterior probability of the non-null models is greater than their combined prior probability.