eXamine: exploring annotated modules in networks

Abstract

Background: Biological networks have a growing importance for the interpretation of high-throughput "omics" data. Integrative network analysis makes use of statistical and combinatorial methods to extract smaller subnetwork modules, and performs enrichment analysis to annotate the modules with ontology terms or other available knowledge. This process results in an annotated module, which retains the original network structure and includes enrichment information as a set system. A major bottleneck is a lack of tools that allow exploring both network structure of extracted modules and its annotations.

Results: This paper presents a visual analysis approach that targets small modules with many set-based annotations, and which displays the annotations as contours on top of a node-link diagram. We introduce an extension of self-organizing maps to lay out nodes, links, and contours in a unified way. An implementation of this approach is freely available as the Cytoscape app eXamine

Conclusions: eXamine accurately conveys small and annotated modules consisting of several dozens of proteins and annotations. We demonstrate that eXamine facilitates the interpretation of integrative network analysis results in a guided case study. This study has resulted in a novel biological insight regarding the virally-encoded G-protein coupled receptor US28.

Figure 1

Data and analysis pipeline. First, control and experimental samples are analyzed to estimate expression levels. Subsequently, gene expression differences (between experiment and control) and their significance are determined. These differences are then mapped to an interaction network, from which a module is extracted with overall significantly-differential gene expression. This module is annotated with overrepresented cell mechanisms from ontology and pathway databases. Finally, the enriched module undergoes iterative visual analysis via eXamine.

Figure 2

Visualization of an annotated module. Interacting proteins with a selection of three subsets, corresponding to overrepresented KEGG pathways. The visualization consists of a combination of a node-link diagram and an Euler diagram.

Figure 3

Comparison. Annotated module visualization using Cytoscape’s Venn and Euler diagram app: (a) Venn diagram and (b) Euler diagram. The number of displayed sets is limited to four and no network structure is shown. (c) Module laid out by one of Cytoscape’s built-in force-directed layout algorithms and BubbleSets superimposed on the network (same color scheme as in Figure 8(b)). Note that it is not immediately apparent that the nodes in the β-catenin set (blue) form a subset of Adherens junction (yellow), because the BubbleSet approach applies no explicit nesting of subsets.

Figure 4

Training neuron n_x,y. (a) The neighborhood within range r_i is trained (colored gray). (b) Certain tiles are already reserved (colored red) in the RSOM algorithm, item t therefore trickles outwards to the best matching free spots (outlined).

Figure 5

Changing the dominance of a set. (a) Highly dominant set, drawing proteins of the set together. (b) Non-dominant set, where the network topology fully defines the layout.

Figure 6

Derivation of contours for set S_i. The darkness of a tile represents the value of the neurons’ i-th component, the thick black line is the contour, dots represent items that are in S_i, and white dots are items that are not in S_i. (a) Contour that results from the union of tiles with a value above a certain threshold. (b) Refined contour with shortcuts across free tiles.

Figure 7

Geometric refinement of set contours after initial layout. Corners are smoothened by dilation and erosion operations, and contours are given a thick and colored internal ribbon. Unique erosion levels create distance between contour outlines, and contour overlap is emphasized by dashed lines.

Figure 8

Item highlighting. (a) Hovered protein (Met) with emphasized interaction links to its neighbors on the right and emphasized sets (KEGG pathways) that contain this protein on the left. Sets outside of the list scope are grouped as markers at the top and bottom, where one set in the bottom group is emphasized. (b) Hovered set (Pathways in cancer) with emphasized member proteins, interactions, and contour.

Figure 9

Case study snapshots. Gene differential expression is shown as a colored box drawn around the node label (green for under-expression and red for over-expression). (a) The annotated module after tagging of the two familiar pathways Pathways in cancer and Phosphatidylinositol signaling system in C1 Two familiar pathways. (b) The annotated module after tagging functions Beta-catenin binding and Growth factor activity in C3 A twist of β -catenin and C4 Growing knowledge. (c) The fully annotated module, including annotation set overview, from which the hypothesis of C5 New insights is derived.

Figure 10

Connection between Met and β -catenin. Proteins that are associated to the selected Adherens junction at the left and corresponding KEGG pathway information at the right, where reactions catalyzed by module proteins are marked in red. Activation of MET by its ligand HGF results in the phosphorylation of β-catenin. This in turn results in its release from cadherin-complexes on the cell membrane into the cytoplasm.