Fig. 1.

Viral community structure. a Non-metric multidimensional scaling (NMDS) of viral community composition in 19 soil viromes obtained from an unsupervised random forest analysis followed by clustering using Ward distances (OOB estimate of error rate = 21.05%). Symbols are color-coded by site (red: North sites; blue: Southeast sites; purple: Southwest sites). b The heat map of contigs differentiating the geographical clusters in a random forest analysis supervised by geographical location and that there is a color gradient which represents the importance. P < 0.01, **; P < 0.05, *