Network Visualization with Cytoscape
https://cytoscape.github.io/cytoscape-tutorials/presentations/network-visualization.html#/
Scientific Visualization Conference
Flatiron Institute
26 October 2017
Goals and Motivations
- Identify relevant types and sources of networks
- Know about various approaches to network visualization
- Know when and how to use Cytoscape
- Publish, share and export networks online
Introductions
- Adjunct Assistant Prof, Pharmaceutical Chemistry
- Executive Director RBVI
- Cytoscape team since 2006
- Author of over a dozen Cytoscape apps
Introductions
- Computational Biologist
- Computational Astrophysicist
- Informatician
- Computer Scientist
- Data Scientist
- Astronomer
- Physicist
- Biologist
- Other?
Introductions
- None -- that's what I'm here for...
- Some network analysis, but what's Cytoscape?
- I've played with Cytoscape, but only for figures
- I do a lot with networks, but mostly in R/Python/Matlab/...
- I do a lot with networks, and use Cytoscape occasionally
- I do a lot with networks, and use Cytoscape a lot
- Lots of both (hey, I could probabaly teach this course!)
Introductions
Caveats
- Cytoscape is primarily used in biology
- My examples (and most of my users) are biologically-related
- I'm most familiar with biological use cases
However...
Introductions
- Cytoscape is not limited to biological use cases
- Some non-biological examples:
- Modeling water systems
- Fictional star distances
- Feel free to ask questions about non-biological use cases!
Introductions
Please install Cytoscape:
Cytoscape 3.6.0 PrereleaseIntroduction to Network Biology
Why Networks?
Networks are everywhere...
especially in biology!
especially in biology!
- Molecular networks
- Cell-cell communication
- Nervous systems
- Social networks
Networks are powerful tools...
especially in biology!
especially in biology!
- Reduce complexity
- More efficient than tables
- Great for data integration
- Intuitive visualization
Networks as Tools
Transcription factors targeting apoptosis genes that bind MKRN1 in protein-mRNA network in embryonic stem cells.
One network figure conveys results from multiple technologies. Cassar, EMBO Reports 2015, Fig.8
Networks as Tools
- Topological properties
- Hubs and subnetworks
- Classify, cluster and diffuse
- Data integration
- Data overlays
- Layouts and animation
- Exploratory analysis
- Context and interpretation
Cassar, EMBO Reports 2015, Fig.8
Applications in Research
Applications in Research
- In 2010, a Nature Methods paper introduced major themes of network visualization.
- (a) Mass spec analysis: 400 PPI in pneumonia microbe.
- (b) Subnetwork with function annotations.
- (c) Layout with knowledge of complexes.
- (d) Collapsed metanodes to reduce complexity.
Gehlenborg (2010) Visualization of omics data for systems biology. Nature Methods, 7, Fig. 1
Applications in Research
- Study of interactions with ASD-associated proteins.
- Y2H with human protein fragments of 26 ASD proteins: 848 interacitons with 539 proteins.
- Merged with known interactions from public datasets.
Sakai (2011) Protein Interactome Reveals Converging Molecular Pathways Among Autism Disorders.
Science Trans. Med., 3(86), Fig. 1
Applications in Research
- SARS-CoV-2-human interactions.
- AP-MS with 26 SARS-CoV-2 proteins: 332 interactions human proteins.
- Merged human-human PPI datasets to identify complexes.
- Used fill color to highlight known drug targets.
Gordon (2020) A SARS-CoV-2 protein interaction map reveals targets for drug repurposing. Nature, 583, Fig. 3
Applications in Research
- Genetic interactions for the entire yeast genome.
- Synthetic lethal screen of 1712 genes: 170,000 interactions.
- Visualization of clustering within and between functional groups annotated by GO.
- Predict functions and identify regulatory subnetworks.
Costanzo (2010) The genetic landscape of a cell. Science, 22, Fig. 1
Applications in Research
- Molecular profiling of two types of brain tumors.
- GSEA to identify distinguishing gene sets (nodes).
- Functional annotation of a network of gene sets.
Witt (2011) Delineation of Two Clinically and Molecularly Distinct Subgroups of Posterior Fossa Ependymoma. Cancer Cell, 20(2), Fig. 4
Applications in Research
- TCGA's seriers of molecular cancer profiles.
- Clinical, genomic, methylation, RNA and proteomic signatures.
- Multiple data types integrated onto signaling network.
- Includes patient sample-level data.
TCGA (2013) Comprehensive molecular characterization of clear cell renal cell carcinoma. Nature, 499, Fig. 4
Applications in Research
Cytoscape
- Open source
- Cross platform
- Consortium
Core Concepts
Networks and Tables
Networks
e.g., PPIs or pathways
Tables
e.g., data or annotations
Visual Styles
Core Concepts
Cytoscape Apps!
apps.cytoscape.org
Loading Network Data
- Cytoscape can import network data from:
- Files (or URLs)
- Excel, TSV, CSV
- XGMML: eXtensible Graph Markup and Modelling Language
- SBML: Systems Biology Markup Language
- BioPAX
- PSI-MI
- SIF: Simple Interaction Format
- GML: Graph Markup Language
- ... and others depending on loaded Apps
- Files (or URLs)
Loading Network Data
- Cytoscape can import network data from:
- Public repositories:
- PSICQUIC
- STRING (via the stringApp)
- IntAct (via the IntActApp)
- Reactome (via the ReactomeFI app)
- WikiPathways (via the WikiPathways app)
- Pathway Commons (via the CyPath2 app)
- NDEx
- Automation:
- Command line scripts
- CyREST via R, Python, etc
- Public repositories:
Loading Table Data
- Cytoscape can load tables from:
- Files (or URLs)
- Excel, TSV, CSV
- Public repositories
- BioMart
- Automation:
- Command line scripts
- CyREST via R, Python, etc
- Files (or URLs)
Saving and Exporting
- Sessions save everything as
.cys files:
Networks, Tables, Styles, Screen sizes, etc. - Export networks in different formats:
SIF, GML, XGMML, BioPAX, JSON - Export tables as CSV files
- Publication quality graphics in several formats:
PDF, SVG, PNG, and JPEG
Network Visualization
Depiction
- Various ways to depict biological networks:
- Node-Link (graph) representation
- Partitioned Node-Link representation
- Matrix representation
- Can be useful for very dense networks
- Can also map information into cells of matrix
- e.g. degree, color scale (heat map)
Data Mapping
- Mapping of data values associated with graph elements onto graph visuals
- Visual attributes
- Node fill color, border color, border width, size, shape, opacity, label
- Edge type, color, width, ending type, ending size, ending color
- Mapping types
- Continuous (numeric values)
- Discrete (categories)
- Passthrough (labels)
Data Mapping
- Avoid cluttering your visualization with too much data
- Highlight meaningful differences
- Avoid confusing the viewer
- Consider creating multiple network images
Layouts
- Layouts determine the location of nodes and (sometimes) the paths of edges
- Types:
- Simple
- Grid
- Hierarchical
- layout data as a tree or hierarchy
- Works best when there are no loops
- Circular (Radial)
- arrange nodes around a circle
- could use node attributes to govern position
- e.g. degree sorted
- Simple
- Types:
Layouts
- Types:
- Force-Directed
- simulate edges as springs
- may be weighted or unweighted
- Combining layouts
- Use a general layout (force directed) for the entire graph, but use hierarchical or radial to focus on a particular portion
- Multi-layer layouts
- Partition graph, layout each partition then layout partitions
- Many, many others
- Force-Directed
Layouts
- Manual tweaks with
Node Layout Tools - Scale
- Align
- Rotate
Layouts
- Use layouts to convey the relationships between the nodes.
- There is not one correct layout. Try different things.
- Layout algorithms may need to be “tuned” to fit your network.
Animation
- Animation is useful to show changes in a network:
- Over a time series
- Over different conditions
- Between species
Analytical Approaches
The levels of organization of complex networks:
- Node degree provides information about single nodes
- Three or more nodes represent a motif
- Larger groups of nodes are called modules or communities
- Hierarchy describes how the various structural elements are combined
Analytical Approaches
Degree is most commonly used, but there are other measures of the relative importance of a single node in a network.
Analytical Approaches
Network topology statistics such as node degree, degree distribution, centralitiy, clustering coefficient, shortest paths, and robustness of the network to the random removal of single nodes are important network characteristics.
Modularity refers to the identification of sub-networks of interconnected nodes that might represent molecules physically or functionally linked that work coordinately to achieve a specific function.
Motif analysis is the identification of small network patterns that are over-represented when compared with a randomized version of the same network. Regulatory elements are often composed of such motifs.
Network alignment and comparison tools can identify similarities between networks and have been used to study evolutionary relationships between protein networks of organisms.
Analytical Approaches
- Concepts
- Graph/Network correspondence
- Node/Vertex correspondence
- Edge directedness
- Usually a network property
- Multigraph
- Allow multiple edges between nodes
- Hypergraph
- Allow edges to connect more than 2 nodes
Analytical Approaches
- Network measures
- Shortest path length
- Shortest traversal distance between two nodes
- Can be weighted if edges have weights or hops if not
- Node degree
- Indegree: # of edges for which this node is a target
- Outdegree: # of edges for which this node is a source
- Degree (undirected): # of edges incident to this node
- Hub
- node with an exceptionally high degree
- Shortest path length
Analytical Approaches
- Network measures
- Diameter: the largest distance between two nodes
- Radius: the minimum of the maximum path between any two nodes
- Characteristic path length: average shortest path length
- Density: extent to which a network is fully connected
- Heterogeneity: tendency of the network to contain hub nodes
- Connected components: number of disconnected groups of nodes
Analytical Approaches
- Network measures
- Clustering coefficient
- Measures how close the neighbors of a node are to being a clique (fully connected)
- Clustering coefficient
Analytical Approaches
- Centrality - measures of node importance
- Degree centrality (find hubs)
- Betweenness centrality (bottleneck)
- Closeness centrality
- Degree centrality (find hubs)
Analytical Approaches
- Scale-free networks
- Degree distribution follows a power law:
P(k) ~ k-y, where y is a constant - Distinct "hubs" (essential proteins?)
- Resilient to perturbation
- Biological (and social) networks tend to be scale-free
- Degree distribution follows a power law:
Analytical Approaches
- Small-world networks
- any two arbitrary nodes are connected by a small number of intermediate edges
- the network has an average shortest path length much smaller than the number of nodes in the network (Watts, Nature, 1998).
- Interaction networks have been shown to be small-world networks (Barabási, Nature Reviews in Genetics, 2004)
Analytical Approaches
- Other techniques
- Guilt-by association
- Combine weak signals to get a stronger one
- Over-representation analysis
- Also called enrichment analysis
- Statistical technique to determine if a set of terms are over-represented (enriched) in a network (or subnetwork)
- Clustering
- Also called unsupervised classification
- Variety of techniques for group nodes based on attributes or connectivity
- Guilt-by association
Analytical Approaches
- Other techniques
- Motif finding
- Finding subnetworks of interest
Cytoscape Apps for Graph Analysis
Thank You!
Network Visualization with Cytoscape
https://cytoscape.github.io/cytoscape-tutorials/presentations/network-visualization.html#/
Scientific Visualization Conference
Flatiron Institute
26 October 2017
Add speaker notes here...