Fig. 5

The modules generated based on networks constructed by different methods on OTU level for the Tara Oceans dataset. a Network constructed by Pearson methods with 100% co-occurrence probability. After performing loose definition of 80% co-occurrence probability and indirect relationships, more OTUs were detected in existing clusters and new clusters are detected for (b) Network constructed by FS-Weight method and (c) Network constructed by PCA-PMI method. d Three closely related clusters detected in the network based on PCA-PMI method. In all four panels, each cluster was composed by phylogenetically close species are labeled in the same color. Based on the calculation of identified OTUs on the family level, the most abundant members of every clusters are annotated