Table 1.
SNP statistics calculated.
Table 2.
Settings for the ABC simulations under scenario 1.
Table 3.
Settings for the ABC simulations under scenario 2.
Fig 1.
Posterior probability in support of the coevolution model (against a neutral model) for scenario 1.
Results shown for 30 repetitions and 30 PODs per value of the cost of infection (s). Results for single PODs are shown as dots. Model choice distinguishing a coevolution model with unknown costs of infection (s), host population size (NH) and parasite population size (NP) from a neutral model with unknown host and parasite population sizes. Note that for these points we added some jitter to the x-values in order to increase the readability of the plots.
Fig 2.
Pairwise Manhattan distance and Δ Tajima’s D (host-parasite) for the PODs under scenario 1 compared to simulations under a neutral model.
Pairwise Manhattan distance (y-axis) and the difference between Tajima’s D of the host and of the parasite (x-axis) for the PODs used for inference in Scenario 1 and the 100,000 neutral simulations run for this scenario. Under the neutral model, host and parasite population sizes vary. Simulations under the neutral model are shown as grey open circles, and a bivariate normal kernel estimation has been applied to obtain a probability density of the summary statistic combinations. The PODs for scenario 1 are shown as diamonds and are coloured coded based on the true cost of infection (s).
Fig 3.
Parameter estimations under scenario 1.
Median of the posterior distribution (y-axis) for the cost of infection s (top, a-c), host population size (NH) (middle, d-f) and parasite population size (NP) (bottom, g-i) when inference is based on host and parasite summary statistics (left), only host summary statistics (middle) or only parasite summary statistics (right). The median of the posterior distribution (after post-rejection adjustment) is plotted for each POD. The true cost of infection for each POD is shown on the x-axis with jitter added to increase the readability. The R2-value of a corresponding linear regression model is shown in each panel.
Fig 4.
Posterior probability in support of the coevolution model (against a neutral model) for scenario 2.
Results are shown for r = 30 and 15 PODs per boxplot. The posterior density in support of the coevolution model (y-axis) is shown for PODs with varying cost of infection (s). The different panels reflect the combination of cH and cP for the respective PODs (left: cH = 0.05, right: cH = 0.1, top: cP = 0.1, bottom: cP = 0.3). Model choice has been run to distinguish a coevolution model with unknown costs of infection (s), cost of resistance (cH) and cost of infectivity (cP) from a neutral model with constant host and parasite population size (NH = NP = 10, 000). Results for single PODs are shown as dots and jitter added to the x-values to increase the readability.
Fig 5.
Pairwise Manhattan distance and Δ Tajima’s D (host-parasite) for the PODs under scenario 2 compared to simulations under a neutral model.
Pairwise Manhattan distance (y-axis) and the difference between Tajima’s D of the host and of the parasite (x-axis) for the PODs used for inference in Scenario 2 and 100,000 neutral simulations. Simulations under the neutral model are shown as grey open circles. A bivariate normal kernel estimation has been applied to obtain a probability density of the different summary statistic combinations. The PODs for scenario 2 are shown in color. Colors reflect the true cost of infection (s) for a particular POD (see legend) and shapes indicate the combination of cH and cP (diamonds: cH = 0.05, cP = 0.1; circles: cH = 0.05, cP = 0.3; crosses: cH = 0.1, cP = 0.1; stars: cH = 0.1, cP = 0.3) for the respective POD.
Fig 6.
Parameter estimations under scenario 2.
Median of the posterior distribution (y-axis) for the cost of infection s (top, a-c), cost of resistance (cH) (middle, d-f) and cost of infectivity (cP) (bottom, g-i) when inference is based on host and parasite summary statistics (left), only host summary statistics (middle) or only parasite summary statistics (right). The median of the posterior distribution (after post-rejection adjustment) is plotted for each POD. The true cost of infection for each POD is shown on the x-axis with jitter added to increase the readability. The R2-value of a corresponding linear regression model is shown in each panel.
Fig 7.
Parameter estimations under scenario 2.
Median of the posterior distribution (y-axis) for the cost of infection s (top, a-c), cost of resistance (cH) (middle, d-f) and cost of infectivity (cP) (bottom, g-i) when inference is based on host and parasite summary statistics (left), only host summary statistics (middle) or only parasite summary statistics (right). The median of the posterior distribution (after post-rejection adjustment) is plotted for each POD. The true cost of infection for each POD is shown on the x-axis with jitter added to increase the readability. The R2-value of a corresponding linear regression model is shown in each panel.