2. Context: Systems story telling
want to help user understand relationships in data so that
they can construct stories about biological systems
How do we use biological information and data to go from a
list of genes to a story about how symptoms of a genetic
disease arise?
[ ]
3. Context: Systems Story Telling
Classification
Reasoning
Model-based
Reasoning
Narrative
Reasoning
(based on B. Mirel, Supporting cognition in systems biology analysis; findings on users’ processes and design
implications. Journal of Biomedical Discovery and Collaboration 2009; 4:1-17)
what of all this
is the right
stuff?
how does it fit
together?
how does it do
something?
what stories
does it tell?
4. Context: Systems Story Telling
Classification
Reasoning
Model-based
Reasoning
Narrative
Reasoning
individual
components
networks/
graphs of
components
system
fragments
“whole”
systems
roles
contextual
classes
Grouping
Patterns
Composition
5. Two kinds of binary relations
individual
components
networks/
graphs of
components
roles
contextual
classes
6. Two kinds of binary relations
component
component
annotation
component
Incidence relation
Graph
,( )
,( )
7. Grouping
Start with
- a network graph of genes, and
- contextual annotations for each
Want to visualize what goes together
[ saw this before in cytoscape.js presentation ]
8. Grouping (hypergraph visualization)
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
GO MF: Voltage-Gated Ion Channel Activity
MeSH: Phosphorylation
MeSH: Signal Transduction
MeSH: Intercellular Signaling Peptides and Proteins
MeSH: Rats
MeSH: Nerve Tissue Proteins
MeSH Ion Channel Gating
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
9. Simplified Groupings
MeSH: Calcium
GO MF: Voltage-Gated Ion Channel Activity
MeSH: Phosphorylation
MeSH: Intercellular Signaling Peptides and Proteins
MeSH: Nerve Tissue Proteins
MeSH Ion Channel Gating
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO MF: Voltage-Gated Ion Channel Activity
MeSH: Phosphorylation
MeSH: Intercellular Signaling Peptides and Proteins
MeSH: Rats
MeSH: Nerve Tissue Proteins
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO MF: Voltage-Gated Ion Channel Activity
MeSH: Signal Transduction
MeSH: Nerve Tissue Proteins
MeSH Ion Channel Gating
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO MF: Voltage-Gated Ion Channel Activity
MeSH: Signal Transduction
MeSH: Rats
MeSH: Nerve Tissue Proteins
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
MeSH: Phosphorylation
MeSH: Intercellular Signaling Peptides and Proteins
MeSH: Nerve Tissue Proteins
MeSH Ion Channel Gating
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
MeSH: Phosphorylation
MeSH: Intercellular Signaling Peptides and Proteins
MeSH: Rats
MeSH: Nerve Tissue Proteins
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
MeSH: Signal Transduction
MeSH: Nerve Tissue Proteins
MeSH Ion Channel Gating
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
MeSH: Signal Transduction
MeSH: Rats
MeSH: Nerve Tissue Proteins
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
computed with BLOSOM (Zhao L, Zaki MJ, Ramakrishnan N., KDD 2006, 827-832), support=5
10. Human-edited Groupings
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
OR
MeSH Ion Channel Gating
MeSH: Phosphorylation
MeSH: Intercellular Signaling Peptides and Proteins
MeSH: Nerve Tissue Proteins
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
MeSH: Calcium
GO BP: Metal Ion Transport
GO MF: Gated Channel Activity
GO MF: Cation Channel Activity
OR
MeSH Ion Channel Gating
MeSH: Signal Transduction
MeSH: Nerve Tissue Proteins
CRKL
ANK3
CABP1
CACNA
2D4
CACNB
3
GNB1
PCBD1
RIMS1
PPM1A
PRKAC
A
RYR2
SRI
CACNA
1C
FAS
FADD
CNTN1
HOOK1
KCNC1
KCNC2
SMAD3
SMAD2
PIK3R1
SPTBN4
SCN1B
SCNN1
B
SCN2A
“user” merged/deleted annotations
11. Formal Concept Analysis
GNB1
PRKACA
RatsCNTN1
PIK3R1
Animals
SMAD2
SMAD3
Calcium
Ankyrins
SPTBN4
CRKL
Cell Line
FADD
Phosphorylation
Ranvier's Nodes
Precipitin Tests
Signal Transduction
Nerve Tissue ProteinsRIMS1
Two-Hybrid System Techniques
Intracellular Signaling Peptides and Pr
• Algebraic approach to
studying data as
incidence relation b/w
objects and attributes
• Construct formal
concepts (biclusters)
• Relate to form formal
concept lattice
12. Formal Concept
CACNA1C
Ion transport protein
KCNC1
KCNC2
RYR2
SCN2A
Beta1 adrenergic receptor signaling pathway
CACNB3
GNB1
PRKACA
Beta2 adrenergic receptor signaling pathway
cation transport
CACNA2D4
SCN1B
SCNN1B
metal ion transport
channel activity
ion channel activity
gated channel activity
cation channel activity
alkali metal ion binding
calcium channel activity
voltage-gated channel activity
transmembrane transporter activity
voltage-gated ion channel activity
substrate specific channel activity
voltage-gated cation channel activity
ion transmembrane transporter activity
substrate-specific transporter activity
cation transmembrane transporter activity
passive transmembrane transporter activity
metal ion transmembrane transporter activity
substrate-specific transmembrane transporter activity
ANK3
RatsCNTN1
PIK3R1
Animals
SMAD2
SMAD3
CABP1
Calcium
SRI
Ankyrins
SPTBN4
CRKL
Cell Line
FADD
Myocardium
Jurkat CellsFAS
PPM1A Smad2 Protein
Phosphorylation
Ranvier's Nodes
Sodium Channels
Calcium Channels
Precipitin Tests
Protein Subunits
Ion Channel Gating
Signal Transduction
Nerve Tissue ProteinsRIMS1
Patch-Clamp Techniques
Sodium Channel Blockers
Calcium Channels, L-Type
Two-Hybrid System Techniques
Ryanodine Receptor Calcium Release Channel
Intracellular Signaling Peptides and Proteins
• Can view incidence
relation as bipartite graph
• A formal concept is a
biclique: a pair of vertices
(A,B) such that for each
object a in A there is an
edge (a,b) for every
attribute b in B
• Can construct as closure
13. Ex: Signal Transduction
Beta1 adrenergic receptor signaling pathway
GNB1
PRKACA
Beta2 adrenergic receptor signaling pathway
Rats
PIK3R1
Animals
SMAD2
SMAD3
Calcium
CRKL
Cell Line
FADD
Myocardium
Jurkat Cells
FAS
PPM1A
Smad2 Protein
Phosphorylation
Calcium Channels
Precipitin Tests
Protein Subunits
Ion Channel Gating
Signal Transduction
Two-Hybrid System Techniques
Ryanodine Receptor Calcium Release Channel
Intracellular Signaling Peptides and Proteins
14. Ex: Signal Transduction
Beta1 adrenergic receptor signaling pathway
GNB1
PRKACA
Beta2 adrenergic receptor signaling pathway
Rats
PIK3R1
Animals
SMAD2
SMAD3
Calcium
CRKL
Cell Line
FADD
Myocardium
Jurkat Cells
FAS
PPM1A
Smad2 Protein
Phosphorylation
Precipitin Tests
Signal Transduction
Two-Hybrid System Techniques
Ryanodine Receptor Calcium Release Channel
Intracellular Signaling Peptides and Proteins
25. Groupings (again)
• A simplified grouping is represented by a minimal
anti-chain in the concept lattice that best covers
objects (e.g., vertices)
• Heuristic:
a. generate concepts for individual annotations
b.for each, create chain by selecting subconcept with
most objects
c. select from chains such that vertex set is covered
32. Patterns
• Previous examples project annotations onto the graph
• For patterns, project the graph onto the annotations
• Can use lattice to find emergent schema (Banks et al.)
• One heuristic: for each edge (a1,a2) find concepts for
a1 and a2, and follow pair of chains upward, adding
edges for new objects, until cannot add more
Banks E, Nabieva E, Chazelle B, Singh M (2008) PLoS Comput Biol 4(10): e1000203
38. More…
• Could keep going …
• Possible to extract association rules, decision trees
and more from concept lattice and use these to
annotate functional relationships in graphs (which
again is something that has been done other ways)
• Also possible to work with multi-valued incidence
relations (e.g., numeric data)
39. Inspired and informed by working with
Barbara Mirel and
NCIBI DBP scientists at U.Michigan
40. Resources
B Ganter, R Wille. Formal Concept Analysis:
Mathematical Foundations, Springer-Verlag, 1999.
!
F Kriegel, Visualization of Conceptual Data with
Methods of Formal Concept Analysis. Thesis,
Technische Universität Dresden, 2012.
!
http://www.upriss.org.uk/fca/fca.html