Table 1.
Summary of ProteinWeaver’s underlying network data. Nodes represent proteins and their encoding genes. PPI: protein-protein interaction. GRN: gene regulatory network. GO: Gene Ontology.
Fig 1.
Network representation and different query examples.
A. Graph representation of Proteins and Gene Ontology terms. Organism-specific networks contain protein-protein interaction edges (black solid lines), regulatory edges (red dashed arrows), and proteins (tan/red circles and blue squares). The GO term hierarchy is represented via the light-blue nodes and directed black solid edges. Proteins directly annotated to GO terms are indicated by blue solid arrows, and blue dashed arrows indicate inferred annotations. B. Different query parameter examples. ProteinWeaver results from a “K Unique Nodes” query connecting flapwing (flw) to k=4 nodes annotated to“myosin binding” (GO:0017022) D. melanogaster with (B.A) physical interactions, (B.B) regulatory interactions, and (B.C) both physical and regulatory interactions. Dashed directed edges indicate regulatory relationships, solid edges indicate undirected physical relationships; see Fig 4B for a full legend describing node and edge types.
Fig 2.
Mixed motifs in ProteinWeaver.
Mixed network motifs identified by ProteinWeaver; adapted from Yeger-Lotem et al. [10]. Black lines indicate physical interactions. Red arrows indicate regulatory interactions. Motifs can contain multiple protein types.
Fig 3.
Proportional distribution of network motifs.
For a detailed breakdown of the motifs by species, see Sect A2.2 in S1 File.
Fig 4.
A) Users enter their protein and GO term of interest, a small integer k, and the species in the top panel and click “Search”. B) The subnetwork is shown via react-cytoscapejs’s plugin. The legend section details the different node and edge types. C) The sidebar includes links to protein and GO term data, basic network statistics, and the ability to export the session. D) The graph exploration section lets users rearrange the network layout and update the query by selecting a new query node or traversing the gene ontology. E-G) The tabbed window shows information about the user-selected protein (including the protein’s rank relative to the GO term), information about the user-selected interaction, and the motifs found in the subnetwork.
Fig 5.
Eb1 to “microtubule bundle formation” (GO:0001578) in D. melanogaster.
(A) “K Unique Paths” mode and (B) “K Unique Nodes” mode with k = 15.
Fig 6.
Gdf6a to “dorsal/ventral pattern formation” (GO:0009953) in D. rerio.
(A) “K Unique Paths” mode and (B) “K Unique Nodes” mode with k = 15.
Fig 7.
ROC curves of four annotation prediction methods.
Receiver Operator Characteristic (ROC) curves of the One-Hop GO Overlap, Hypergeometric Distribution, Degree, and RandomWalk annotation prediction methods for all species. The ROC AUC is reported in the legend.
Fig 8.
Precision-Recall curves of four annotation prediction methods.
Precision-Recall curves of the One-Hop GO Overlap, Hypergeometric Distribution, Degree, and RandomWalk annotation prediction methods for all species. The PR AUC is reported in the legend.