Table 1.
Summary of different studies, parasite-vector combinations and experimental proceedures used to parameterise the model.
Fig 1.
Structure of mSOS: a multiscale model of the population dynamics of Plasmodium falciparum during sporogony.
(A) Multiple malaria parasites are found within a single mosquito; we separately model development time from inoculation at blood feeding (G) to oocyst (O) and from oocyst to salivary gland sporozoites (S). If dissected, a mosquito is “positive” for a particular parasite life stage if at least a single parasite has developed. We do not model the decline in observed oocyst numbers due to oocyst bursting, since we do not have sufficient later oocyst observations. (B) Mosquitoes are infected via a membrane feeder; parasite load varies in each due to differences in the number of parasites ingested and variation in mosquito immune response. (C) Temporal dynamics of sporozoite prevalence within a mosquito population: following the infectious blood feed, a proportion of the population is infected with malaria parasites. The parasites develop into sporozoites causing the mosquitoes to become infectious. Throughout the experiments, mosquito mortality (in the laboratory) may be greater in infected mosquitoes, resulting in an eventual decline in observed sporozoite prevalence.
Fig 2.
Single temperature 27°C model fit to the oocyst and sporozoite data.
The panels show our model fit to the 27°C dataset: panel A to the oocyst prevalence, panel B to the oocyst intensity data and panel C to the sporozoite prevalence. (A & C) points show parasite prevalence of the laboratory mosquito data (95% confidence intervals are given by the point range). The grey shaded area represents the 95% credible interval of the model posterior predictive means, the median posterior predictive mean is shown by the black line. (B) The points show the mean parasite load among all blood fed mosquitoes (intensity); the point range indicates the 2.5%–97.5% quantiles of the raw data. The shaded area represents the 2.5%–97.5% quantiles of the negative binomial distribution; where the location and overdispersion parameters are set to their posterior means.
Fig 3.
Model fits to sporozoite prevalence data across all temperatures.
These fits were generated by fitting a single model to all temperatures simultaneously (“all temperature” model), with the functional form of temperature as described in Eqs (2.11) and (2.12). Black points: parasite prevalence of the laboratory mosquito data (95% confidence intervals are given by the vertical black lines). The grey shaded area represents the 95% quantiles of the posterior predictive means; the black lines represent the median posterior predictive means.
Fig 4.
Effect of temperature on malaria transmission parameters.
Panel A shows the model impact of temperature on the EIP quantiles (as indicated in legend); panel B shows its impact on the human-to-mosquito transmission probability. In both panels, the lines show impact as estimated by the all temperature mSOS model with 95% posterior intervals indicated by shading; the discrete round points show the independent estimates from the single temperature mSOS model at each temperature, 95% posterior credible intervals are shown by the vertical lines. The discrete triangle points show the logistic growth model parameter estimates at each temperature. The number of iterations used to calculate the EIP plot were thinned to every 5th iteration for efficiency.
Fig 5.
Modelled impact of parasite load on the extrinsic incubation period.
Panel A shows the impact of varying the mean parasite load of infected mosquitoes on the temporal dynamics of sporozoite prevalence in a sensitivity analysis; panel B summarises how the EIP is affected by the same parameter in the sensitivity analysis. All other parameters were held constant at their mean posterior values.