Fig 1.
Estimates of relatedness with 95% confidence intervals.
Estimates and confidence intervals are shown for all 325 choose two (52650) P. falciparum sample pairs and are ordered by increasing relatedness estimate. Confidence intervals are coloured according to classifications based on lower and upper confidence interval end-points, where τ is an arbitrary threshold used to classify highly-related pairs. For example, a pair is considered highly-related with τ = 0.25 if the lower end-point of the confidence interval around its relatedness estimate exceeds 0.25. Otherwise stated, if its relatedness estimate is statistically distinguishable from 0.25.
Table 1.
Classification of parasite sample pairs.
Classification is based on the lower and upper end-points (LCI and UCI, respectively) of the 95% confidence interval around each relatedness estimate, , where ϵ is an arbitrarily small number to identify LCI ≈ 0 and UCI ≈ 1 given that LCI and UCI ∈ (0, 1) not [0, 1]; and τ is an arbitrary threshold used to classify highly-related pairs. We use ϵ = 0.01 throughout, τ = 0.25 (main analysis) and τ ∈ {0.25, 0.50} (sensitivity analysis).
Fig 2.
Fractions of highly-related sample pairs partitioned in time and space.
(A) Partitioned by time between collection dates. (B) Partitioned by collection city, where the inter-city great-circle distance is the distance in kilometres (km) between city pairs on the Earth’s surface.
Fig 3.
P. falciparum population connectivity assessed using a threshold-free approach.
1-Wasserstein distance between parasite populations from different cities versus inter-city great-circle distance in kilometres (km).
Fig 4.
P. falciparum population connectivity based on fractions of highly-related sample pairs.
The width of each inter-city edge is proportional to the fraction of highly-related sample pairs across cities plotted in Fig 2(B). Note that the edges between Guapi and Quibdó and Guapi and Tadó are plotted but too thin to visually discern.
Fig 5.
Clonal components and the average relatedness between them.
Vertices depict clonal components, which are groups of two or more statistically indistinguishable parasite samples. CC vertices are plotted using the Fruchterman-Reingold layout algorithm [45], thereby clustering inter-related CCs. The size of each CC vertex is proportional to the number of parasite samples per CC, ranging from 2 to 28 statistically indistinguishable parasite samples. CCs are named in order of the collection date of the earliest parasite sample per CC (S2 Table). CCs with parasite samples collected from two or more cities are depicted as pie charts. Colour denotes the city of parasite sample collection. Edge transparency and weight is proportional to average relatedness, ranging from 0.003 to 0.840. Relatedness estimates that are indistinguishable from zero were set to zero. Edges whose average relatedness is zero are not plotted. Each CC besides CC4 is related to at least one other. CC4 contains two samples (one from Tadó, another from Quibdó). It is likely a contaminant; see main text.