Typically, we don't have enough time to discuss all of the capabilities of Cytoscape visualizations.
This module tries to fill in some of the gaps. Here are the topics we're going to cover:
Ring charts provide ability visualise top-5 terms in the network using split-charts. You can change
the number of terms and type of chart in chart setting panel.
Click Draw Chart to create chart and split charts
will appear in the network.
To change charts setting, go to Visit the network specific chart settings
Provides the option to create groups from network clusters
Working with Groups in Cytoscape
Group visualization
None: no visualization for the group. Grouped nodes look the same as other nodes.
Compound Node: visualize the group as a compound node, essentially a node
(the group) that contains other nodes (the members of the group). Individual member nodes can be individually
selected and manipulated.
Show Group Node: the group node itself is added to the network and "membership edges" are added between the group node and each group member
Single Node: similar to compound node in that all of the member nodes
are visible, but member nodes can not be individually selected or manipulated. Essentially, the
entire group is treated as a single node.
Working with Groups in Cytoscape
Cytoscape groups can optionally aggregate member node data.
Aggregations can occur on all columns, or only specified columns
Aggregation options:
Numeric columns: None,
Average, Minimum value,
Maximum value,
and Sum
String columns: None,
Comma-separated Values,
Tab-separated Values,
Most Common Value, and
Unique Values
Boolean columns:
None,
Logical AND, and
Logical OR
List columns:
For numeric lists, same options as numeric columns plus Concatenate and Unique Values
For String lists, None, Unique Values and Concatenate
Working with Groups in Cytoscape
Configuring Groups
Groups are configured using the Group Preferences dialog (Edit→Preferences→Group Preferences)
Working with Groups in Cytoscape
Try it out!
Open up Cytoscape, and select the Ivacaftor Coauthor network. Play around with grouping either by
creating groups manually or by installing the setsApp and creating groups based on the node type column.
Taken from 10 simple rules to create biological network figures for
communication (PLoS Comp Bio 2019)
Ten Simple Rules
Rule 1: Determine first the network specifics: purpose, data, scale, salient aspects
Must decide:
the purpose of the figure
the data components of the network
any additional constraints on the figure
the relative weight of these requirements
What is the story you want to tell? What is your evidence to support it?
Ten Simple Rules
Rule 2: Depending on scale, consider alternative layouts to node-link diagrams
Node-link diagrams aren't always the best representation for a network
In particular, with lots of nodes and edges, they tend to turn into "hairballs"
Alternative representations include:
matrices
hive plots
biofabric
Ten Simple Rules
Rule 3: Beware of unintended spatial interpretations
Whether intended or not, viewers will interpret spatial relationships:
proximity: things that are closer are more similar/related
centrality: things in the center are more relevant than those on the periphery
direction: up is good, down is bad, and information flows horizontally (left to right in western cultures)
Ten Simple Rules
Rule 4: Provide readable labels and captions
Any text in the figure should support the purpose of the figure (see Rule #1)
Not all nodes/edges need to be labeled. If it can't be read, it isn't useful.
Rule of thumb: Font sizes for labels should be no smaller than the caption font.
Labels should be properly placed in close proximity to the subject.
Ten Simple Rules
Rule 5: Consider aggregation to show only what is required; choose the right level of detail
One way to reduce the complexity of a network is to use aggregation. For example:
Group very similar nodes together as a single "metanode"
Group nodes based on co-location
Combine members of a complex into a single node
Ceate multiple figures showing different node sets
Use dimensionality reduction (e.g. PCA) to remove nodes
Ten Simple Rules
Rule 6: Use color responsibly
Color is a big topic, too complex to address completely now. However, some key ponts
are:
Color is perceived, so might be slightly different for each individual
Color blindness effects a significant (8%) population, so avoid red-green color scales
Intensity (luminance) is more readily perceived than hue or saturation (chromaticity).
Saturated colors draw attention
Palettes have been designed to "do the right thing" -- use them (e.g. Brewer color palettes, Viridis, etc.).
Three different types of palettes:
Divergent: color gradients with a natural midpoint (e.g. log2 fold change values)
Qualitative: categorical data (e.g. protein family)
Sequential: color gradients from low to high values (e.g. node degree)
Ten Simple Rules
Rule 7: Use other visual marks and channels appropriately: dots and arrows, size, shape, position
Marks are geographical elements (dots, shapes, and lines)
Dots and glyphs should be used to represent items
lines and arrows should be used to represent relationships
Blobs are used to represent regions or grouping
Channels define how those are rendered (position, color, shape, size, tilt, area, and volume).
Position, color, and shape are used for identity or categorization
Size, area, color (with caveats -- see rule #6), and volume are used to distinguish quantities or attribute strength
Channels can be combined to overcome encoding limitations (e.g. color "red" + shape "stop sign")
Ten Simple Rules
Rule 8: Use layering and separation
The goal is to increase the prominence of the important information by imagining it as sitting on a higher layer. We do this by:
Adjusting the weight, luminance, saturation, opacity, or size of the layers (e.g. decrease the luminance and opacity of the unimportant parts of the figure)
Using spatial cues (e.g. grouping) to highlight relationships
Ten Simple Rules
Rule 9: Use multiple figures instead of one
Sometimes its difficult to adhere to all of the above rules in a single figure. So, use multiple figures!
Ten Simple Rules
Rule 10: Do not use unjustified 3D
Three dimensions can be very helpful for perceiving data which corresponds directly to 3-dimensional objects
(animals, organs, molecules, physical space), but can be distracting and difficult to perceive correctly when
applied to other data sets
Data Visualization with Cytoscape
Slides: http://tutorials.cytoscape.org
John "Scooter" Morris and Alex Pico
Online
25 October 2024
Add speaker notes here...