MOSAIC

GO network annotation and partition in Cytoscape

Mosaic performs network annotation and interactive partitioning driven by the Gene Ontology. The Mosaic algorithm works by first annotating the network with GO terms, followed by partitioning the network into a series of subnetworks based on the biological process annotation of nodes.
Mosaic was created as a Google Summer of Code project by Chao Zhang in collaboration with the National Resource for Network Biology (NRNB). Please contact us with any feedback or questions.

And if you end up using Mosaic in your research, please cite this publication:
Chao Z, Hanspers K, Kuchinsky A, Salomonis N, Xu D, and Pico AR. Mosaic: Making Biological Sense of Complex Networks. Bioinformatics 2012, 28, 14:1943-4.

Mosaic is supported by the National Resource for Network Biology (P41 GM103504).

To install, place the following jar files in the plugins directory of your Cytoscape installation:

Mosaic is an open source project. Feel free to contact us with any questions.

The Mosaic algorithm workflow is explained in the below graphic. Click to enlarge.

1. Select a network to load under File>Import Network (Multiple File Types). The network can be in any of the Cytoscape supported network file formats.

Download sample file: mosaic-demo-yeast.sif (14KB)

1. Start Mosaic by selecting Plugins>Mosaic.

2. If this is the first time running Mosaic on your network, you will see a message in the Retrieve GO Annotations section of the Mosaic interface to download necessary databases. Click Download to continue. Mosaic will download the necessary database for your data. It is only necessary to download datbases once per species.
3. Once database download is complete, Mosaic will prompt you to annotate your network with GO annotation. Click Annotate to continue. You can save the annotated network as a session file for future use. You can also re-annotate the network with the latest annotations from GO at any time by clicking Re-annotate.

1. In the Select Attributes section, verify that the correct attributes are selected.

   Note: It is possible to select node attributes other than GO annotation for partitiong and node color, but not for layout. Using alternate annotations involves importing annotation data prior to running Mosaic.

2. In the Set Parameters section you can choose several parameters that will directly affect the results of Mosaic:
   • Minimum nodes to view: The minimum number of nodes in a subnetwork. Any subnetworks with fewer nodes will not be shown. Default is 5.
   • Maximum nodes to view: The maximum number of nodes in a subnetwork. Any subnetworks with more nodes will not be shown. Default is 200.
   • GO level cutoff for partition: The level of the GO hierarchy used for partitioning into subnetworks. Selecting a deeper level (larger number) results in a larger number of subnetworks. You can always drill deeper into a given subnetwork after running the initial partition. Default is 2. Note that for the purposes of this manual, GO level cutoff was set for "All the way", meaning to the deepest possible level of GO.
   • Spacing between nodes: Spacing between nodes in a given subnetwork. Default is 30.
3. Click Run to start Mosaic.

When partitioning is complete, subnetworks will be displayed as tiles as shown below. In addition to subnetworks representing GO biological process terms, the display also includes an overview network (top left tile) and the original network (bottom right tile).

In the Mosaic tab in the left-side Control Panel, all subnetworks will be listed, including subnetworks that fall outside defined thresholds.

• To locate a particular subnetwork, select the network in the Mosaic side panel to highlight it in the tiled view.
• To view a particular subnetwork, click the Maximize button at the top left of the tiled window.

Each subnetwork represents a biological process and the layout is controlled by cellular component annotation using a pre-set template. Node coloring is determined by molecular function annotation, by default.

The network can be manipulated like any other Cytoscape network:

• Move nodes by click and drag to optimize the layout.
• Delete nodes and/or edges.
• Zoom in and out.
• Use the Network Overview Pane in the lower left corner of the Cytoscape window to navigate around the network.
• Highlight a set of nodes to view associated attributes in the Data Panel. To add specific attributes to the display, click the Select Attributes button in the Data Panel and select the attributes you want to see from the list.
• Search for specific nodes by typing the name of the node in the Search field in the Cytoscape toolbar. If the node is found, this will refocus the network view on the node.

It is also possible to highlight nodes in a selected subnetwork based on molecular function annotation, using the Node color selection menu at the top of the Mosaic tab. With a subnetwork selected, select a molecular function from the drop-down and all nodes with that annotation will be highlighted. To display the highlighted nodes in the Data Panel, click the Select nodes button.

It is possible to partition a select branch of the subnetwork tree after an initial analysis is completed. This works on any non-end branch whos sub-branches did not result in any subnetworks in the initial partition. For example, if none of the branches possible subnetworks fall within the node thresholds, the subnetworks will not be automatically created, but these could still be paritioned manually using this process.

1. Select a non-end branch by clicking on it in the Mosaic tab in the Control Panel.
2. Right-click on the branch and select Partition This Network. You will be able to select a specific subnetwork to create, or you can select all.

Note that Mosaic will use the same parameters for paritioning as were selected in the original analysis.

In addition to the subnetworks, an overview network is also created in which each subnetwork is represented as a node. The size of each node represents the number of nodes in that subnetwork, and the line thickness represents the number of connections between two subnetworks. You can adjust the edge widths using Viz Mapper.

In addition to the visual style that is applied to the overview network, the corresponding attribute values used to create the visual style is available as a set of node and edge attributes. The attributes are

• evidences: nodes classified in the GO/process (node attribute)
• subnetworkSize: number of nodes classified in the GO/process(node attribute)
• commonEvidences: nodes in common (edge attribute)
• overlapCount: number of nodes in common (edge attribute)

Annotations other than GO can be used for partitioning and node coloring in Mosaic. The annotation data can be imported as node attributes. The below instructions assume that the annotation data is available in a simple text file, and that the primary identifier in the annotation file is the same as that used for nodes in the network. This example explains how to use information on predicted miRNA binding site for node coloring.

1. After importing a network, go to File>Import>Attribute from Table (Text/MS Excel)... to open the Import Annotation File interface.
2. Select the annotation file. In the Advanced section, select Show Text File Import Options. In the Attribute Names section, select Transfer first line as attribute names.
3. In Advanced, select Show Mapping Options and verify that the correct column in the annotation file is mapped to the appropriate network attribute (i.e. matching identifiers).
4. In the Preview section, deselct columns that you don't want to import by clicking on the column header. The column should be highlighted in grey if de-selected. The interface should now look like this:
5. Click Import to start the import.
6. When import is finished, select a few nodes in the network and verify in the Data Table that the annotation was imported correctly.
7. Start Mosaic. In the Select Attributes section, select the imported attribute under Attribute to use for node coloring. Click Annotate to annotate the network with GO for partitioning and layout.
8. Each subnetwork will now be colored based on the predicted miRNA binding sites.

If these descriptions do not help address a problem you are experiencing, please contact us.

Make sure you only have one copy of Mosaic and CyThesaurus installed.

If you install using the Cytoscape Plugin Manager and also download and install the jars from this website, then there will be conflicts and odd behaviors. This is true for any Cytoscape plugin. You can remove manually installed plugins from the plugin folder in your Cytoscape installation; and you can remove managed plugins from Plugins>Plugins Manager.

Mosaic can not annotate yeast networks using SGD IDs, like S000006332.

Mosaic isn''t able to translate SGD IDs to Ensembl, so the annotation of networks with these IDs fails. We will fix address this in a future release. In the meantime, work with another identifier (e.g., YPR128C).

Error message: "Please load a network first!"

You must have at least one network loaded in Cytoscape before running Mosaic.

Error message: "CyThesaurus 1.31 or later verion is necessary for runing Mosaic!"

CyThesaurus 1.31 or later version is required by Mosaic for annotating networks and earlier versions might cause inconsistent issues. You can get the latest version at the top of this page. Be sure to uninstall older versions first.

Error message: "You need internet connection for downloading databases."

Mosaic supports many species and cannot distributed them all inside plugin. So, the first time you work with a given species, an internet connection is required to download the species-specific database files. Subsequent runs for this species will work offline.

Error message: "Building *** subnetworks may take a while. Are you sure you want to proceed?"

The message means that the current network contains more than 100 subnetworks and partitioning procedure will take longer. You can decline and then try decreasing the level to which to partition or try reannotating with GOslim (if you were working with GOfull).

Error message: "Selected attribute does not match layout template".

This message appears when the full GO hierarchy is used to annotate nodes, as opposed to Mosaic-specific GO slim. If your network is already annotated with GO terms, make sure to re-annotate the network in Mosaic, to ensure that the correct GO annotations are used.