Modelling malaria pathogenesis

Abstract

Almost 20 years after the development of models of malaria pathogenesis began, we are beyond the 'proof-of-concept' phase and these models are no longer abstract mathematical exercises. They have refined our knowledge of within-host processes, and have brought insights that could not easily have been obtained from experimentation alone. There is much potential that remains to be realized, however, both in terms of informing the design of interventions and health policy, and in terms of addressing lingering questions about the basic biology of malaria. Recent research has begun to iterate theory and data in a much more comprehensive way, and the use of statistical techniques for model fitting and comparison offers a promising approach for providing a quantitative understanding of the pathogenesis of such a complex disease.

Fig. 1

Schematic of two recent models of malaria pathogenesis. A. Modified from Mideo et al. (2008), this model tracks the densities of red blood cells (RBCs), merozoites and gametocytes. The main regulatory mechanism here is resource (i.e. RBC) abundance. B. The model of Dietz et al. (2006) focuses on the effects of innate and acquired immune responses and tracks the density of infected RBCs only. The abundance and action of different immune effectors is translated into probabilities of infected RBCs surviving their attack.

Fig. 2

Model selection and validation. Data from a single CD4⁺ T cell-depleted mouse (dashed lines and dots) and predictions from four models (solid lines) Top panels, RBC densities; bottom panels, parasite densities. Model predictions are from four models representing different hypothesis about what regulates the dynamics of pathogenesis: i. no RBC age structure or parasite cell age preference and constant erythropoetic response; ii. no RBC age structure or parasite cell age preference and variable erythropoetic response; iii. RBC age structure, possible parasite cell age preference and constant erythropoetic response; iv. RBC age structure, possible parasite cell age preference and variable erythropoetic response. Models iii and iv provide statistically significantly better fits to the RBC data than models i and ii. As the models were fit only to the RBC data, the parasite data provide a means of model validation. It is clear that model iv is better than iii at qualitatively capturing the parasite dynamics. Model iv is selected as the ‘best’ model among those tested. See Mideo et al. (2008) for further details.