Identifying spatio-temporal dynamics of Ebola in Sierra Leone using virus genomes

Abstract

Containing the recent West African outbreak of Ebola virus (EBOV) required the deployment of substantial global resources. Despite recent progress in analysing and modelling EBOV epidemiological data, a complete characterization of the spatio-temporal spread of Ebola cases remains a challenge. In this work, we offer a novel perspective on the EBOV epidemic in Sierra Leone that uses individual virus genome sequences to inform population-level, spatial models. Calibrated to phylogenetic linkages of virus genomes, these spatial models provide unique insight into the disease mobility of EBOV in Sierra Leone without the need for human mobility data. Consistent with other investigations, our results show that the spread of EBOV during the beginning and middle portions of the epidemic strongly depended on the size of and distance between populations. Our phylodynamic analysis also revealed a change in model preference towards a spatial model with power-law characteristics in the latter portion of the epidemic, correlated with the timing of major intervention campaigns. More generally, we believe this framework, pairing molecular diagnostics with a dynamic model selection procedure, has the potential to be a powerful forecasting tool along with offering operationally relevant guidance for surveillance and sampling strategies during an epidemic.

Keywords: Ebola; disease modelling; phylodynamics; spatial epidemiology.

Figure 1.

Virus genome data from EBOV cases in Sierra Leone characterizes the spatial spread of the epidemic. (a) The time course is shown for the number of confirmed cases [40] and sequenced EBOV genomes [33]. Three stages of the epidemic are highlighted. (b) Genetic linkages are illustrated with ancestors (open circles) and descendants (closed dots), both coloured by the origin district shown in the map key. The blue arrow highlights a linkage from the Western Urban to Kenema districts. (c) Chiefdom populations (greyscale) and major roads (yellow traces) are illustrated on the map of Sierra Leone. The blue arrow highlights the fastest driving route between the Western Urban to Kenema district. (d) All transmission distances are shown in a cCDF. The distribution of transmission distances are fit by a power law with ρ = 1.66. The blue arrow follows the linkage from (b) and (c). (e) Two spatial models are plotted as maps representing the probability of observing a new case linked to the Western Urban district, using (ρ^⋆ = 1, τ^⋆₂ = 1) for the gravity model and ρ = 1.66 for the power law. (f) The log-likelihood ratio, , comparing the gravity and power-law models, is plotted for 50-day windows. The dashed black line represents (ρ = 1.66, τ₂ = 1) fixed in time; the solid black line of uses the MLE (ρ^⋆(t), τ^⋆₂(t)), computed for each window. The solid red trace describes the number of linkages.

Figure 2.

The empirical power law for the transmission distances. (a) The complementary cumulative distribution function (cCDF) for Stage I, (50 ≤ t < 200 days), is plotted along with the power-law model using the MLE value of ρ* = 1.8 ± 0.1 for N = 54 linkages. (b) The cCDF for Stage II, (200 ≤ t < 350 days), is plotted with the power-law model using the MLE value of ρ = 1.6 ± 0.1 for N = 332 linkages.

Figure 3.

The partially observed transmission network (POTN) and estimation of parameters for the gravity model. (a)–(c) The left column illustrates the POTN for Stages I–III. The open circles and closed dots represent ancestors and descendants, respectively. Each are coloured by the ancestor district. Linkages of shortest duration are shown. The right column illustrates the POTN linkages on a map with black arrows highlighted in yellow. The black arrow width is proportional to the number of linkages. (d) The likelihood evaluation is illustrated for each stage along a grid of values (ρ, τ₂). The MLE and 95% CI are illustrated by a red dot and black ellipse, respectively.