idatavisualizationlab/biolinker

BioLinker: Bottom-up Exploration of Protein Interaction Networks

Please click to watch the overview video.

Please find our manuscript submitted to the 6th IEEE Symposium on Biological Data Visualization (BioVis 2016) here.

Try BioLinker Web demo. https://idatavisualizationlab.github.io/BioLinker/talks/BioLinker.html

BioLinker is an interactive visualization system that helps users to perform bottom-up exploration of complex protein interaction networks. Five interconnected views provide the user with a range of ways to explore pathway data: Overview/ protein selector, context view, main view, publication view, and conflict matrix.

1) Overview / Protein Selector:

This panel provides an overview of a subset of millions of index cards in the database, such as protein interaction within the cos-7 cell line. Users can select any protein within this overview network to start with. Users also have the option to instead input protein name into a search box as depicted in the left panel of the following figure. This will perform a request to load the selected protein and its immediate neighbors from our index card database. As users iteratively expand the subnetwork in the main view, the overview keeps track of the expanded sub-network over the overall context as depicted in the right panel.

2) Main View:

In this view, node (protein) sizes are computed based on the number of direct neighbors. Edges (index cards) are color-encoded by interaction types. BioLinker supports finding paths between selected proteins. The following figure shows an example. Users specify source, target, and the maximum number of hops in between source and target. BioLinker displays all possible paths under that condition. Source node is pinned to the left while target node is pinned to the right of the visualization. The shortest path from PIK3CA to TRAF6 goes through two hops Akt and NF-kappaB. In this example, we also overlay cancer genomics data onto the network: purple nodes are proteins with high copy number alteration in the Bladder Urothelial Carcinoma study (TCGA, Nature 2014). BioLinker accesses this cancer study on cBioPortal through its web service interface.

3) Context View:

Systems biologists and cancer researchers are frequently interested in understanding the contexts of biochemical reactions and comparing protein interaction sub-networks by context. The top left panel in the next figure shows the context view for the network in the main view (top right panel). In particular, we show a 2-degree separation network of a selected protein antigen which is located in the center of the main view. The lower panel depicts brushing and linking of two views. We have selected mouse in the species category. Other context categories are updated accordingly. In the main view (on the bottom right), we notice that all protein interactions in mouse are between antigen and its immediate neighbors, but not the second degree separated neighbors.

4) Publication View:

We use TimeArcs visualization to show the discoveries of these index cards by publication year. A request to load relevant publications for new index cards is sent to the PMC publication database on our server every time the protein network is expanded. The next Fig.(b) shows an example of publication view of the graph in Fig.(a). In particular, time axis goes from left (2004) to right (2012). An arc connects two proteins/complexes at a particular time (based on when the interaction was discovered/ publication year). The colors encode interaction types, such as green for increase_activity, red for decrease_activity, orange for translocation, and blue for binds. As depicted in Fig.(c), mousing over an arcs display publication data associated to an index card, such as paper title, authors' names, authors' contacts, affiliations, publication year, journal name, and external link. Users can go to the actual paper by a clicking on the provided link.

5) Conflict Matrix:

We use an comparator script to detect potential conflicts between index cards in the main view. For example, we select TGF-beta as protein of interest and iteratively expand the network around this protein. The index card comparison results are presented in form of an adjacency matrix. In each cell, we draw an arc symbol for each interaction (colored by type) of the two participating index cards. Cells with orange background contain two index cards with potential conflict as depicted in the left panel of Fig.(a). In Fig.(b), we inspect a potential conflict cell in the matrix (left). Two corresponding index cards are highlighted in the publication view (right). We can see that the two index cards have the same participants (TGF-beta and Smad7), but opposite interaction types (increase_activity versus decrease_activity). This indicates conflicting knowledge obtained from two different publications in 2010 and 2011. We can look further into details of publication data for each index card to verify these conflicting information as shown in Fig.(c).

Acknowledgments

This work was funded by the DARPA Big Mechanism Program under ARO contract WF911NF-14-1-0395.