Incorporating genetic selection into individual-based models (IBMs) of malaria and other infectious diseases.

Abstract

Optimal control strategies for human infections are often investigated by computational approaches using individual-based models (IBMs). These typically track humans and evaluate the impact of control interventions in terms of human deaths, clinical cases averted, interruption of transmission etc. Genetic selection can be incorporated into these IBMs and used to track the spread of mutations whose origin and spread are often driven by the intervention, and which subsequently undermine the control strategy; for example, mutations which encode antimicrobial drug resistance or diagnosis- or vaccine-escape phenotypes. Basic population genetic descriptions of selection are based on infinite population sizes (so that chance fluctuations in allele frequency are absent) but IBMs track finite population sizes. We describe how the finite sizes of IBMs affect simulating the dynamics of genetic selection and how best to incorporate genetic selection into these models. We use the OpenMalaria IBM of malaria as an example, but the same principles apply to IBMs of other diseases. We identify four strategies to incorporate selection into IBMs and make the following four recommendations. Firstly, calculate and report the selection coefficients, s, of the advantageous allele as the key genetic parameter. Secondly, use these values of s to calculate the wait-time until a mutation successful establishes itself in the population. The wait time for the mutation can be added to speed of selection, s, to calculate when the mutation will reach significant, operationally important levels. Thirdly, quantify the ability of the IBM to robustly estimate small selection coefficients. Fourthly, optimise computational efficacy: when s is small it is plausible that fewer replicates of larger IBMs will be more efficient than a larger number of replicates of smaller size.