Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
Interactive Network Construction
Using R and cRytoscape
Paul Shannon
Institute for Systems Biology
August 25th, 2009
Interactive Construction: begin, add,
reduce
• Your experimental data
• Add annotation
• Calculate functional enrichment
• Connect nodes (by function, known interactions,
pathway membership)
• Eliminate uninteresting edges and nodes
• Design the next experiment
33 Early Spike Genes
Bioconductor Graph + Reworked
Cytoscape 1.1: cRytoscape
> g = new ('graphNEL', edgemode='directed’)
> nodeDataDefaults (g, ’geneSymbol') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', ’geneSymbol') = “KLF3”
> nodeLabel (g, ’geneSymbol'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘geneSymbol’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, ’geneSymbol') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', ’geneSymbol') = “KLF3”
> nodeLabel (g, ’geneSymbol'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘geneSymbol’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, ’geneSymbol') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', ’geneSymbol') = “KLF3”
> nodeLabel (g, ’geneSymbol'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘geneSymbol’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, ’geneSymbol') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', ’geneSymbol') = “KLF3”
> nodeLabel (g, ’geneSymbol'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘geneSymbol’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, ’geneSymbol') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', ’geneSymbol') = “KLF3”
> nodeLabel (g, ’geneSymbol'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘geneSymbol’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, 'label') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', 'label') = “KLF3”
> nodeLabel (g, 'label'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘label’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, 'label') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', 'label') = “KLF3”
> nodeLabel (g, 'label'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘label’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
cRytoscape Commands
> nodeDataDefaults (g, 'label') = '’
> g = addNode (‘51724’, g)
> nodeData (g, '51274', 'label') = “KLF3”
> nodeLabel (g, 'label'); render ()
> broadcast (c (‘crytoscape.cmd’, ‘vizrule’, ‘nodeLabel’,
‘lookup’, ‘label’))
> nodeFontSize (20); render ()
> broadcast (c (‘crytoscape.cmd’, "nodeFontSize", 20))
Early Spike Gene Symbols
GO BP for Early Spike Genes
Rendered
An aside: rules for node color & size
nodeColor (g.go, 'score',
c ('below:white',
'0:white',
'200:red',
'above:blue'));
nodeSize (g.go, 'count',
c ('below:30',
'1:30',
'10:100',
'above:120'))
Interpolating Network Edges
(HPRD ‘empirical’)
5 Interpolated Edges
Profile of Interpolated Genes: NFKB1 &
HMGB1
Reject or Tentatively Accept?
HMGB1 & FOXC1: abstract ('11748221’): HMGB1 interacts with many apparently
unrelated proteins by recognizing short amino acid sequences … [has] the ability to bind
DNA nonspecifically and to interact with various proteins … transcriptional activators,
repressors, co-repressors
HMGB1 & NFKB1: abstract ('12604365'): HMGB1 interacts differentially with members of
the Rel family of transcription factors ... enhances DNA binding by p65/p50 and p50/p50,
but reduces binding by p65/p65 ... Functionally, HMGB1 is required for the
NF-kappaB-dependent expression of the adhesion molecule VCAM-1.
Reject or Tentatively Accept?
NFKB1 & ATF3: abstract ('7692236'): Cyclic AMP-independent ATF family members
interact with NF-kappa B and function in the activation of the E-selectin
promoter in response to cytokines….
A ‘Little Language’ for Network
Visualization from R
defaultEdgeStyle (graph, style)
defaultNodeBorderColor (graph, color)
defaultNodeBorderStyle (graph, style)
defaultNodeColor (graph, color)
defaultNodeShape (graph, shape)
defaultNodeSize (graph, size)
defaultSourceArrow (graph, arrow)
defaultTargetArrow (graph, arrow)
destroy ()
eda (graph, edge.attribute.name)
eda.names (graph)
edgeColor (graph, edgeAttributeName, keyValuePairs)
edgeColorDiscrete (graph, edgeAttributeName, keyValuePairs)
edgeStyle (graph, edgeAttributeName, keyValuePairs)
format.long.label (labelString)
hide ()
invert ()
A ‘Little Language’ cont.
layout ()
noa (graph, node.attribute.name)
noa.names (graph)
nodeBorderColor (graph, nodeAttributeName, keyValuePairs)
nodeBorderStyle (graph, nodeAttributeName, keyValuePairs)
nodeColor (graph, nodeAttributeName, keyValuePairs)
nodeColorDiscrete (graph, nodeAttributeName, keyValuePairs)
nodeFontSize (newSize)
nodeLabel (graph, nodeAttributeName)
nodeShape (graph, nodeAttributeName, keyValuePairs)
nodeSize (graph, nodeAttributeName, keyValuePairs)
render ()
select (node.names)
sourceArrow (graph, edgeAttributeName, keyValuePairs)
targetArrow (graph, edgeAttributeName, keyValuePairs)
unhide ()
Conclusions
• The interpretation of experimental data is inherently exploratory.
• A huge amount of annotation, interaction and pathway information
is available, some of it useful in that exploration.
• Point-and-click analysis tools are good, but …
• Exploration via scripting can sometimes take you further
• R and bioconductor offer scripting and many data packages
• cRytoscape adds a network visualiztion tool to R
• Not yet ready for release, but if anyone is interested, contact me.