Figure 1.
The relative proportions (%) of various data source types that were used are shown categorized according to IMG/M microbiome taxa at the class level. Panel A shows the proportions (%) with respect to the total number of data sets while Panel B shows the proportions (%) with respect to the total number of genes.
Figure 2.
Network Construction Workflow.
Operonic genes are predicted on the basis of co-direction and intergenic proximity using scaffolds containing more than one gene (Panel A). Operons and their constituent genes can be filtered according to the presence or absence of a target annotation such that at least one member of an operon is required to possess a target descriptor (Panel B). Note that the filter step is optional and can applied to obtain target perspective networks while being omitted in the construction of source perspective networks. Each gene in a given operon is mined for its various types of functional annotations where any particular type has a domain of existing values (Panel C). For each operon, the obtained functional annotations are used to infer bidirectional functional interactions for annotations having the same type but different values (Panel D). Note that interactions are inferred directly for immediately adjacent gene pairs and also transitively for downstream members within the same operon.
Figure 3.
Target Stringency versus Network Coverage.
Four polyketide target perspective networks were constructed with progressively increasing target stringency and each network was translated into each of the four annotation categories. The proportion of nodes and edges in each polyketide network was compared to its corresponding overall network. Panel A shows that coverage for nodes decreased for all annotation categories with increasing target stringency and Panel B shows that coverage for edges also decreased for all annotation categories with increasing target stringency.
Table 1.
Summary of Network Features.
Figure 4.
Metagenomic Cellulase Networks.
The target perspective networks for cellulase functional interactions are shown where large node diameter represents high node degree within each respective network. Panel A shows a network constructed using MetaCyc annotations with a highly connected central hub having the annotation PWY-1001: cellulose biosynthesis. The highlighted nodes represent the top ranking module which is enlarged in Panel B. Panel C shows a network constructed using COG annotations and features a highly connected central hub with the annotation COG1363: cellulase M and related proteins. The highlighted nodes represent the top ranking module which is enlarged in Panel D.
Table 2.
Top Ranked Functional Modules.
Figure 5.
The annotative themes for the top ranked MetaCyc module are depicted where the numeric values indicate the number of annotations belonging to a thematic category. Specifically, amino acid categories are represented vertically and metabolic categories are represented horizontally. Note, the vertical themes are encapsulatory while the horizontal themes are mutually exclusive. A variety of functional perspectives can be simultaneously visualized by way of the interacting and overlapping thematic sets.
Figure 6.
The source perspective networks for human gut functional interactions are shown where large node diameter represents high node degree within each respective network. Panel A shows a network constructed using KEGG annotations where the highlighted nodes represent the top ranking module which is enlarged in Panel B. Panel C shows a network constructed using TIGRFAM annotations where the highlighted nodes represent the top ranking module which is enlarged in Panel D.
Figure 7.
The comparative networks for human gut functional interactions are shown where large node diameter represents high node degree within each respective network. Specifically, two networks were constructed using TIGRFAM annotations and compared for mutual versus exclusive nodes. Panel A shows the intersection of the networks where the highlighted nodes represent the top ranking module which is enlarged in Panel B. Panel C shows the difference of the networks where the highlighted nodes represent the top ranking module which is enlarged in Panel D.