Fig. 4: Strain-sharing and social network position.
From: Gut microbiome strain-sharing within isolated village social networks
a, The strain-sharing rate by geodesic distance is shown. Null distributions were calculated based on 10,000 random samples from a within-village microbiome permutation. The null distribution slopes downwards because of the effect of large networks; larger villages have more pairs with a higher geodesic distance between them and, on average, a lower average strain-sharing rate between individual people (see Supplementary Fig. 10 for more details); 95% confidence intervals are plotted around the null distributions. People have more similar microbiomes to their first- and second-degree social connections that expected due to chance. b, As a person’s degree centrality (number of social connections) increases and they are more socially embedded in the village, their average strain-sharing rate with the village also tends to increase (Pearson correlation ρ = 0.16; P = 1.54 × 10−12 two-sided correlation test). The grey shaded area around the regression line represents the 95% confidence interval. c, Example social network for Hondarribia. People with more connections (degree) and a higher network centrality (betweenness or eigenvector) tend to be more microbially typical. Nodes are coloured according to their average strain-sharing rate with the rest of the village, and nodes are sized according to their degree centrality. d, When people have a wider variety of social connections (increasing degree centrality), between-host heterogeneity tends to increase, and people on average have a lower strain-sharing rate with their first-degree connections (Pearson correlation ρ = −0.18; P = 3.6 × 10−14 two-sided correlation test). The grey shaded area around the regression line represents the 95% confidence interval. e, Example social network for Hondarribia. People in the centre, with a wider variety of social connections, have, on average, a lower strain-sharing rate with their first-degree connections. Nodes are coloured according to average strain-sharing rate with first-degree connections, and nodes are sized according to degree centrality.
