What are the main features of OmicsNet?

OmicsNet has been developed to allow researchers for creation and visualization of relationships among genes, proteins, miRNAs, transcription factors, and metabolites in a three-dimensional (3D) space. In particular:

1. From a list (or multiple lists) of molecules of interest (genes, transcription factor, miRNA or metabolites) to molecular interaction networks using build-in comprehensive knowledgebase.
2. Visual exploratory analysis of the network in our innovative 3D network visualization system with rich features for layout, coloring, highlighting, functional and topological analysis.

Some example visualization results can be seen from our Overview page.

What are the data formats accepted by OmicsNet?

1. Molecular Lists: OmicsNet accepts one or multiple lists of genes, transcription factors, miRNA and metabolites. The format is such that the first column takes the input type and the second column takes the expression value or any other quantitative measurement. Multiple ID types are supported for different organisms.
• List of 691 DE genes from helminth infection signature (download)
• Data from multi-omics (transcriptomics, metabolomics) study to identify biomarker of intrahepatic cholangiocarcinoma cancer (ICC)
• List of gene symbols (download)
• List of metabolites (download)
2. Network Files: three different formats are supported including .sif, .txt and .graphml format. Please click on the following links to see example files supported:
   • Example .sif format
   • Example .txt format
   • Example .graphml format

Which browsers are supported by OmicsNet?

The 3D visualization system was developed based on the Web Graphics Library or WebGL technology. WebGL is the standard 3D graphics API for the web. It allows developers to harness the full power of the computer’s 3D rendering hardware from within the browser using JavaScript. Before WebGL, developers had to rely on plugins or native applications and ask their users to download and install custom software in order to deliver a hardware-accelerated 3D experience.

WebGL is supported by most major modern browsers that support HTML5. We have tested OmicsNet in several major browsers (see below). Our empirical testings have shown that Google Chrome usually gives the best performance for the same computer:

What if WebGL is supported but disabled on my browser?

How are the networks generated from my data?

OmicsNet build interaction networks starting from list(s) of input molecules: gene/protein, TF, miRNA and metabolite and supports the integration of up to three different interaction types. Using the list of biomolecules, OmicsNet searches for directly interacting partner in the interaction database selected to build the interaction network. It is required to specify the order of network creation: primary, secondary or tertiary.Four basic rules dictate network creation:

1. Primary interaction network is composed of seeds and its immediate interacting partners.
2. Secondary and tertiary interaction other than PPI will query for direct interactions against gene/proteins contained in the primary network. This procedure will add new nodes and edges in the network.
3. PPI as secondary or tertiary interaction will query for interactions among the current gene/proteins in the network (adding edges only).
4. If more than one input list is uploaded, the lists of secondary and/or tertiary interaction serve as a constraint to limit the search space of interactions.

Case 1: upload a list of genes/proteins and search against our databases for a mix of regulating actors (TF/miRNA) or metabolites. This allows the construction of a co-regulatory network or metabolic network. An example of network construction starting from a gene list is shown below:

If the user input is a list of miRNA, TF or metabolites. These queries will be mapped to the selected database and a search is performed to identify proteins directly interacting with these queries. The network contains both the queries and genes/proteins interacting with it. After this initial step, the resulting interacting genes/proteins are used to search for other interaction types.

Case 2: starting from on a TF list and identify which genes are regulated by them (TF-gene as interaction 1). This TF-regulatory network can be further enriched by adding PPI interaction (interaction 2) between the regulated nodes. In this case, setting PPI as secondary interaction will only introduce new edges among current gene/protein nodes in the network.

If the users upload more than one type of molecules, the list of molecules of secondary and tertiary interactions serve as limit the search space.

Case 3: starting from more than one type of molecules. In this case, it is advised to set PPI or metabolite as primary interaction as the confidence of prior-knowledge for these interaction data is higher than the other types. The additional uploaded lists will then serve as constraint to limit the search space.

The above approaches will typically return one giant subnetwork ("continent") with multiple smaller ones ("islands"). Most subsequent analyses are performed on the continent. Note, networks with less than 3 nodes will be excluded.

Why adding PPI as secondary/tertiary interactions will only add edges?

This process adds an extra layer of information while constraining the network size and maintaining the focus on the nodes of interest (first interaction). In our interaction models, the gene/proteins act as common factor (hub) that links the other interaction types, therefore all primary interaction network contains gene/proteins. Adding PPI interaction will resulting in adding edges between existing protein nodes in the current network.

Which databases are used for OmicsNet?

• PPI: STRING database (2017), IntAct database (2018), InnateDB (2015)
• TF-gene regulation: JASPAR TF-target from Harmonizome (2017), ENCODE Chip-seq data (2017), TRRUST (2017)
• miRNA-gene regulation: miRBase(2017), TarBase(2017)
• Protein-metabolite interaction: KEGG(2018), Recon 2 (2017)

What are some user cases of network building in OmicsNet?

Please consult the tutorials for more details. Below are some examples of biological questions that OmicsNet can address.

1. Determine which miRNAs target a list of gene:
List of gene/proteins and select miRNA-gene regulation as primary interaction.
2. Identify the direct interacting partners of a list of gene/protein:
List of gene/proteins and select PPI interaction as primary interaction.
3. Identify the enzymes interacting with a list of metabolites
List of metabolites and select enzyme-metabolite as primary interaction.
4. Identify TFs and miRNA targeting input gene/proteins and its direct interacting proteins.
List of gene/proteins, select PPI interaction as primary interaction and select TF-gene and miRNA-gene as secondary/tertiary interaction.
5. Identify interactions between input gene/proteins + direct interacting proteins and input TF.
List of gene/proteins and TF, select PPI interaction as primary interaction and select TF-gene as secondary interaction.
6. Identify interactions between input gene/proteins + direct interacting proteins and input TF/miRNA.
List of gene/proteins, TF and miRNA, select PPI interaction as primary interaction and select TF-gene and miRNA as secondary/tertiary interaction.

Which databases are used for creating the PPI network?

OmicsNet uses high-quality protein-protein interaction (PPI) data from STRING. database, the most comprehensive PPI database to date. This database collects and integrates already published experimental data on PPI in addition to predicted interactions. A filter can be applied to subset higher confidence interactions based on STRING provided combined score of confidence and whether the interactions have experimental evidence. Our human PPI database contains around 670 000 interactions while our mouse database contains around 900 000 interactions. To supplement human and mouse data, we also provide PPI data from InnateDB, a database built by manual curation from published literature as well as experimental data from several PPI databases including IntAct, MINT, DIP, BIND and BioGRID.

Which databases are used for transcription factor-gene regulatory data?

TRRUST: Manually curated transcription factor targets database, derived from ~12,000 pubmed articles describing small-scale transcription regulation experiments.

JASPAR: Database of transcription factor binding motifs. The TF-gene information is obtained by using position weight matrices of transcription factors binding preferences are used to scan the promoters of all genes present in the range of 500 bp upstream to 2000 bp downstream of a transcription factors start site. For a gene to be considered as targeted by a transcription factor, a 100% match to consensus sequences of the promoter must be obtained through the scan. Mice data are derived from human data.

ENCODE: Transcription factor target data derived from ENCODE Chip-seq data. We performed an initial filter to keep peak signal with score above 500 only and performed TF targets prediction on the binding data using BETA minus algorithm. It outputs a regulatory potential score based on the over-representation and distances between the binding sites and gene transcription start site within a range of genomic region. An average of the regulatory scores are taken for targets if more than one experiment is done on the same TF. We applied a filter to only include TF targets having a regulatory score more than 1. Please refer to BETA paper for more information on gene matching. Mice data are derived from original experiments performed on mouse tissues or cells.

Wang, Su, et al. "Target analysis by integration of transcriptome and ChIP-seq data with BETA." Nature protocols 8.12 (2013): 2502-2515.

Which database are used for miRNA-protein regulatory data?

The data is from miRNet which has collected data mainly from miRTarBase and TarBase. Both databases curate their experimentally validated microRNA-target interactions from the literature. For humans, we have around 330 000 interactions involving 2596 different miRNAs. For mice, there are around 58 000 interactions involving 769 miRNAs.

Which databases are used for protein-metabolite interaction data?

Metabolite-protein interaction data were obtained from Recon 2 for human. Recon 2 is one of the most complete metabolic reconstruction of human metabolism. This model was built by a community of domain experts which aggregates information from cell type/organ specific metabolic reconstructions such as HepatoNet1, literature and previous knowledge bases such as Recon 1 and EHMN (Edinburgh Human Metabolic Network).

KEGG: species-specific metabolite-protein interaction data derived from KEGG reference reactions by mapping these reference reactions to species-specific and reaction-specific enzymatic proteins and metabolites.

How many genes can be visualized (network size limit)?

The visualization is limited by the performance of users' computers and screen resolutions. Too many nodes and edges will result in greater latency in manipulating the network and more severe "hairball" effect. Based on empircal tests, we have set the maximum number of nodes to 10000. We recommend limiting the total number of nodes to between 200 ~ 2000 for best experience. For very large networks, please make sure you have a decent computer equipped with a performant graphics card.

How do I reduce the network size?

OmicsNet offers various options for reducing the network size.

1. To compute the minimum network connecting all the seed entities. In filter menu, select "Minimum Network".
2. To filter the networks to remove undesired nodes based on betweenness and degrees.
3. To delete promiscuous nodes
4. To use "Extract Module" function located in the bottom of toolbar to build a subnetwork using an arbitary list of nodes in the current network.
5. To reduce the input genes by using larger fold change and/or smaller p value cutoffs;

How do I improve the visualization of larger networks?

To deal with "hairball effect" associated with larger networks, OmicsNet offers several options that can partially address this problem.

1. Reducing the opacity of edges reduces edge occlusion problem. In Edge Style drop-down menu, select Opacity option.
2. Use edge-bunding algorithm to deform and aggregate closely located edges to provide a less cluttered view of the network. It is also located in the Edge Style menu.
3. If it is a composite network(i.e. multiple types of nodes), apply the 2D perspective multi-layered view to separate molecular types by layer. Click on icon located in the vertical toolbar.

How to iteratively add nodes in the current network (targeted integration)?

Instead of creating composite network from the network building page, the users can perform targeted integration in the network visualization page. This procedure consists of searching for biomolecules that have targets in the current network (using Regulation Explorer) after which the users select the ones they are interested to add in the current network.

How do I integrate expression profile into the network?

When uploading list of molecules, an optional second column can be included to add expression value information. In the network visualization page, display the expression profile by clicking on "Expression" in View drop-down menu.

How do I display different topological properties on the "Node Explorer" table?

Go to "Node Attribute" submenu located in "Style" menu and click on the desired property to be displayed in the table.

What is a module and how are modules identified in OmicsNet ?

The connections between nodes of a biological network are not random. Modules refers to parts of the network that have higher connections within than the average connections expected across the whole network. It is especially an important aspect of biological network as it can help in identify nodes that are likely to participate in the same biological function as they are more likely to be tightly connected as a subgroup of the network.

OmicsNet currently uses three different module detection algorithms including Walktrap, InfoMap and Label Propagation. WalkTrap exploits the idea that random walking tends to remain in a dense part of network which corresponds to a module. InfoMap algorithm joins neighboring nodes into modules and these modules are further merged into supermodules using the map equation principle. The map equation minimizes the steps needed for random walker to describe a network by finding network modules. Label Propagation works by labeling nodes with unique labels at the initial step and, subsequently, nodes take the label that most of its neighbors have until a consensus label is reached for all nodes. There is no clear advantage on using one algorithm over another, users are welcome to try all three.

How do I interpret the differences in node colors and sizes in the default network?

In the network generated by OmicsNet, the size of the nodes are based on their degree values, with a bigger size indicating larger degrees. The color indicates the type of nodes.

How do I change the network layout?

To change the layout, please go to the menu bar located on top of network viewer and click on "Layout" dropdown menu to show options including defaut force-directed, hierarchical, layered and spherical layouts. The layered mode improves the visualization of composite networks where there are multiple interaction types.

What is Force-Module layout?

It is a variant of standard force directed layout in which graph communities the nodes from a same graph community are more tightly clustered together forming clusters of nodes. To distinct visually the graph communities, nodes from the same community are colored the same and a transluscent spheres are added. Double-clicking an individual module or selecting it within module table will put it into the view focus.

How do I change the color of modules?

Click on the color cell located at the right-most side of the result table from "Module Explorer" and choose a color from the color picker.

Can I change node color or size?

You can change the color and size of a node. The shape cannot be changed in the current implementation. To change the node color, you can use our "Coloring Option" panel to change node color based on its source molecule type. To change node size, you can keep clicking it (double-clicking) to increase its size. You can also use the Node Size functions to increase or decrease size of all nodes.

To modify an individual node, you can use our "Modify Node" panel located at the bottom left size of the window. You can either manually enter the ID of the node to be modified or, on the graph, right click on the node of interest and click on "Modify Node" option from the context menu.

How do I color the nodes based on different network topological properties?

The nodes can be colored based on degree, betweenness, centrality, expression (user-supplied), etc. To do so, click on Global button in Coloring Options panel located on the top left corner. In the opened window, select an attribute for which the nodes will be colored in the drop-down menu. A selection of color schemes are offered from which users can choose from as well.

How do I scale the node sizes based on different network topological properties?

The node sizes can be scaled based on degree, betweenness, centrality, expression (user-supplied), etc. The default node sizes are scaled by node degrees. To do so, click on Size option from the Node Style drop down menu. In the opened window, select an attribute in the drop-down menu from which the size will be scaled to.

Can I change edge color, width and opacity?

To change edge color, opacity and width, users can go to Edge dropdown menu located on top menu bar.

Can I change the background color of the network?

Yes, OmicsNet allows users to freely choose the background color. Users can click on the background of "Coloring Options" panel located on the top left corner of the screen or go to Style menu to select the color..

How do I change the position of a single node and group of nodes?

To drag a single node, simply hold down left click boutton and drag and drop the node to new position. To drag a group of nodes (2D viewer only), make sure to highlight the group of nodes that are to be moved then in "Select" menu and go to Scope submenu to select the option Highlighted. Then drag and drop one of the highlighted genes to the new position.

How can I label the nodes in the network?

Right click node and select Add Label option from the context menu.

Nodes will be automatically labeled when their sizes reach a certain threshold. Therefore, you can simply increase node size to label any node. To do so:

• Label a single node: right click a node and choose "Add Label" option in the context menu.
• Label all highlighted nodes: use the Node tab in the Display Options panel on top right, select "Highlighted nodes" and "Increase ++", then keep clicking Submit button to increase the size until labels show up.

How do I manually highlight any arbitrary nodes of the network?

To do this, you can use the "Batch Selection" panel on the right side of the page. You need to supply a list of the nodes of your interest.

How do I differentiate seed nodes from other nodes?

To highlight nodes are from input list, please click on highlight seed nodes icon located in the vertical tool bar located on the top left corner of Network Viewer. To change the halo color, use the color picker located on the top of tool bar.

How do I use halo highlight effect for all the subsequent selections?

The default behavior of OmicsNet is highlighting by color for the selection of genes from enriched pathways, detected modules and double click. To use halo highlight effect instead, click on More Options and go to Highlight tab to select Halo Effect Only.

Can I extract a module or a highlighted section from the network?

Yes. To do this, first select or highlight section of the network, then go to Module Extraction tab in Advance Options (3D Viewer) or click the Extract icon on the left tool bar in the network view of window (2D Viewer). The operation is expensive, and you have to wait a few seconds for the extracted network to return. The returned network will be named as "moduleX" and is available in the "Network Explorer" panel on the top-left of the page for future reference.

Note: 2D viewer is more adapted for manual selection of nodes for the purpose of node extraction. 3D depth perspective make node selection a tedious process.

How does the extract module function work?

This process is similar to the Steiner Tree problem where the algorithm identifies a minimal subnetwork containing all the terminal nodes (highlighted nodes) from the complete network (the current subnetwork). We implemented a heurisitc approach that provides an approximate answer to this problem to reduce computation time.

How do I turn off the fog effect?

By default, there is a fog effect in the scene to accentuate the depth effect of 3D network. To turn off this effect, click on Fog located on the vertical toolbar.

What is edge bundling function?

Node-link diagram often suffers from "hairball effect" visual clutters due to large amount of nodes and edges.Edge bundling improved the readability of such network by using a process analogous to bundling network cable wires. To perform edge bundling click on located in the vertical toolbar.

How to export an animation of the current network?

Select GIF animation option from Download submenu.

Can I perform GO or pathway analysis on my significant genes alone?

Yes, you can test enriched gene ontologies or pathways (KEGG/Reactome) for only your query genes. To do so, first select and highlight query genes using the Highlight Color toolbar on the top left (you may have to highlight twice for upregulated and downregulated genes respectively); or you can use the Hub Explorer and select queries from the node table. After that, select a functional catergory from the Function Explorer section, and click the Submit button.

Can I test the enriched functions of my highlighted selection?

Yes. Users can perform enrichment tests on currently highlighted nodes in the network. There are two different ways to highlight nodes

• Module highlight: perform module detection and select a module on the table.
• Manual highlight: select nodes from the data table on the left or by double clicking nodes. Select a "Include eighbours" highlight scope in "Select" menu to to highlight node and its immediate neighbours.

How is enrichment analysis performed?

The enrichment analysis is to test whether any functional modules (gene sets) from the user selected library are significantly enriched among the currently highlighted nodes within the network. OmicsNet uses hypergeometric tests to compute the enrichment p values. Note that in the current implementation, enrichment analysis only takes into account gene/proteins in the network, not the other types of molecules.

What is the difference between "Enrichment Analysis" module and "Functional Explorer" module?

Both approaches uses Over Representation Analysis (ORA) approach, however the input genes are different in each case. "Enrichment Analysis" uses all genes or metabolites uploaded by the user while "Functional Explorer" uses genes present in the current interaction network. ORA on interaction network (Functional Explorer) has the effect of amplifying biological signal by including direct interacting partners of seed molecules in the analysis.

How do I perform integrative transcriptomics and metabolomics enrichment analysis?

First, upload both gene and metabolite lists. In the network building page, select "KEGG projection" option under "Metabolite-Protein Interaction" and press proceed. This option projects the input data into a global metabolic network built from KEGG. In the side panel, integrative enrichment analysis can be performed. Two different statistical integrations are provided:

• Direct merging: ORA performed on reference pathways contain both gene and metabolite
• Combining P-value: ORA performed on gene-specific and metabolite-specific reference pathways followed by combing P-values using Stouffer's method. Weight is calculated from the percentage of input features mapped to features contained in reference pathways.