Symbolic Systems Technology - csl.sri.com · Symbolic Systems Technology Using formal systems and...
Transcript of Symbolic Systems Technology - csl.sri.com · Symbolic Systems Technology Using formal systems and...
Symbolic Systems Technology
Using formal systems and and reasoning tools to• Understand how things work; why things don’t work• Design, analyze, adapt complex systems
Carolyn TalcottISSSB
2011 November 14
Plan
• What I mean by Symbolic Systems Biology
• Pathway Logic as an example
• Challenges & Opportunities
Elements of Symbolic Systems Biology
• Modeling and analysis of biological systems
• Representation in a formal system / logic
• Executable models
• Use of formal reasoning tools to answer questions
Modeling Issues
• What questions do you want the model answer?
• What can you observe/measure?
• What questions do you really want the model answer?
• What does that mean?
• Explain it to a computer!
Modeling Methodology
Impact
rapid prototyping
state space search
S |=! model
checking
What is a formal system?
• Language: to describe things and properties
• Semantics: thing satisfies property
• Reasoning principles: proving/disproving properties of things
• [Reflection: to model and reason about models and reasoning]
• Executable formal models (ala model train, airplane, ...)
• System state: collections of entities (name,location,knowledge,resources..)
• System behavior: transition rules, response to stimuli
• Execution: application of rules, possibly concurrent--watch it run, po
• Executions are objects that we can reason about
• Properties of states (P,Q) and executions (ϕ: P until Q, eventually P, this event depents on that event )
Symbolic analysis -- answering questions
• Forward collection -- upper bound on possible states
• Backward collection -- initial states leading to states of interest
• Search -- for states of interest
• Model checking -- do all executions satisfy ϕ, find counter example
• Constraint solving -- steady state analysis
Model Transformations: new views, faster analysis
• Mapping of formal systems
• access to tools (RWL to Petri Nets)
• alternative semantics (qualitative, kinetic, probabilistic/stochastic)
• richer property specification (RWL to HOL)
• Abstracting
• quantities (points to regions)
• qualities (phos(Y 123) vs phosY vs phos vs act/inact)
• entities (protein vs family or homology class)
• Canonical form -- to combine/compare models
Pathway Logic (PL) http:pl.csl.sri.com
•Essence•Examples• Response to Stimuli• Bacterial Protease Network• TB metabolic network • Sleep
In a nutshell
• Represented in rewriting logic, mapped to Petri nets for model checking
• Queryable formal knowledge base of experimental findings (Datum KB)
• subject, assay, treatment, times, cells, result, variants, source
• Algebraic signature ~ controlled vocabulary linked to standard sources (Uniprot ...).
• Terms represent reference entity with modifications, activity state, location
• Global model: network of executable rules (RKB) inferred from Datum KB
• Local Model ~ initial state (experimental setup) + relevant rules from RKB
• Pathways (sets of biomolecules that function together) and regulation effects discovered by asking questions (answered by formal reasoning tools)
Answering questions using PL Assistant
• Specify goals and avoids
• Find subnet
• backward collection followed by forward collection
• Find pathway
• assert goal not achievable and ask LoLA, pathway is counter example
• In silico knock outs
• Comparing networks
• Following connections
• Find all rule minimal pathways (separate tool, to be integrated)
• cluster
• ask for essential rules/occurrences, uses
Signaling response to Egf (AMK)
Two ways to activate Erk
Datum query
V9V7V8
V12
V10
V11
V22
V23
V19
V21
V20
V24
V40
V41
V37
V39
V38
V42
V28
V29
V25
V27
V26
V30
V46
V47
V43
V45
V44
V48V3V1V2V6V4V5
V16
V17
V13
V15
V14
V18
V34
V35
V31
V33
V32
V36
V78
V77
V76
V74
V75
V50
V68
V73
V72
V71
V69
V70
V67
V66
V65
V63
V64
V49
V57
V62
V61
V60
V58
V59
V112
V113
V109
V111
V110
V114
V142
V143
V139
V141
V140
V144
V106
V107
V103
V105
V104
V108
V136
V137
V133
V135
V134
V138
V100
V101
V97
V99
V98
V102
V130
V131
V127
V129
V128
V132
V94
V95
V91
V93
V92
V96
V124
V125
V121
V123
V122
V126
V54
V55
V51
V53
V52
V56
V82
V83
V79
V81
V80
V84
V88
V89
V85
V87
V86
V90
V118
V119
V115
V117
V116
V120
050
100
150
200
Cluster Dendrogram
hclust (*, "ward")
dataDist
Height
All ways to activate Erk (144), clustered
Essential rules cluster 1 (green)
RapA-inhib-CLc
81
MurAA-CLc
45
MurAA-degraded-CLc
CiaRH-CLc
142140
40
MecA-degraded-CLc
101
Mpr-gene-on-CLc
99
Bpr-gene-on-CLc
97
NprE-gene-on-CLc
181
SpoIVFB-inhib-CLm
RapA-CLc
80
BofA-CLm
103
NprB-gene-on-CLc
SpoIVFB-CLm
182180
proPhrA-XOut
112114
PhrA-XOut CSF-XOut
115
PhrG-CLc
120
CtsR-degraded-CLc
43
DegS-CLc
122
DegU-phos-CLcCtsR-CLc
68 52
50
51
FtsH-gene-downreg-CLc
RapE-inhib-CLc
PhrE-CLc
193
30
ClpE:ClpP-CLcRapF-inhib-CLc
PhrF-CLc
194
ClpE-CLc
64
RapA-gene-CLc
155
RapA-gene-downreg-CLc
31
ClpC:ClpP-CLc
RapG-CLc
121
33
109
Epr-gene-on-CLc
RapG-inhib-CLc
ClpP-gene-on-CLc
SpoJ-gene-on-CLc
SpaA-CLw
23 22
Heme:IsdG-CLc
18
113
PhrA-CLc
SigmaA-CLc
61 606291
58
SipV-gene-downreg-CLc
ClpC-gene-on-CLc
35394142 38 37189
44
47 48 49 46
17
Fe3+-CLc
124
SipT-gene-on-CLc
56
SipT-gene-downreg-CLc
Heme:IsdI-CLc
RapE-CLc
79
HtrA-gene-on-CLc
SipV-gene-CLc
57
SipS-gene-downreg-CLc
Abh-CLc
126
SipS-gene-on-CLc
Abh-gene-on-CLc
136
LytE-degraded-CLw
173
LytE-CLw
174 171
LytF-degraded-CLw
59
Lsp-gene-downreg-CLc pre-AmyQ-CLm
130129 128
FtsZ-CLc
158160 163 161 159 162
LytF-CLw
172
Lsp-gene-CLc
167
LonB-gene-on-CLc
SigmaB-inhib-CLcFtsZ-inhib-CLc
165
SigmaH-degraded-CLc
29
ClpP:ClpX-CLc
36
164
ClpX-CLc
54
EzrA-CLm
157
LonA-CLc
166
SigmaG-inhib-CLc
93
Spo0A-CLc
SigmaB-CLc
63 67 176175178179
89
Spo0B-phos-CLc
Spo0F-CLc
90
Spo0B-CLcSpo0A-phos-CLc
ClpX-gene-downreg-CLc
55
73
SpoIIAA-gene-on-CLc
ClpX-gene-CLc
SpaA-anchored-CLw
24
SrtF-CLm
EzrA-degraded-CLm
SigmaW-CLc
2527 32
FosB-gene-on-CLcFtsH-gene-on-CLc
SecY-CLm
71
MecA-CLcSecY-degraded-CLm
pre-PrsA-CLm
131
ComA-CLc
146
75
Spo0A-gene-on-CLc
PrsA-CLm
138
156
CodW:CodX-CLc
Spo0E-degraded-CLc
SrfA-gene-on-CLc
AmyQ-XOut
Lsp-CLm
105
Vpr-gene-on-CLc
SigmaH-CLc
pbpE-gene-on-CLc
107
WprA-gene-on-CLc
SigmaH-act-CLc
CodW-CLcCodX-CLc
72
SpoIIE-gene-on-CLc
186
SpoIIAB-gene-on-CLc
190
SpoIIAA-CLc
SpoIIAA-phos-CLc
183
SpoIVFA-cleaved-CLm
SpoIVB-CLc
Vpr-XOut
116
Vpr-gene-CLc
147
SrfA-gene-CLc
SrfA-CLc
106
145
143
SpoJ-gene-CLcSpoJ-CLc
108
SigmaK-CLc
pro-SigmaK-CLm
Heme:IsdA-anchored-CLw
13
DegU-degraded-CLc
YwbN-gene-CLc
133
134
YwbN-CLc
KinE-CLc
88
28
PBP4-CLmWprA-gene-CLc
KipI-inhib-CLc
WprA-CLw
Spo0F-phos-CLc
YwbN-gene-on-CLc
84
KipA-CLc
Lipid-II-bound-CLw
Lipid-II-CLm
21
SigmaG-CLc
pbpE-gene-CLc
70
9
Heme:IsdC-anchored-CLw
IsdH-anchored-CLw
ComK-degraded-CLc
ComS-CLc
IsdA-anchored-CLw
10
IsdB-anchored-CLw
11
7
ComS-degraded-CLc
AbrB-gene-downreg-CLc
111 95 96
135
AbrB-gene-CLc
AbrB-CLc
119
AprE-gene-downreg-CLc ClpC-gene-downreg-CLc ClpP-gene-downreg-CLc
McsA-CLc
53
Spx-CLc
McsB-CLc
Spx-degraded-CLc
SigmaX-CLc
132
195
SinR-CLc
118123
AprE-gene-on-CLc
proCSF-XOut
Parks-nucleotide-CLc
19
Undecaprenyl-phosphate-CLm
Lipid-I-CLm
20
MraY-CLm
SpaB-CLwSpaC-CLw
SpaABC-pilus-Sig
Hpr-CLc
117
RapF-CLc
SpoIVFA-CLm
184
151
ComK-gene-on-CLc
CtpB-CLc
153
DegU-CLc
150
ComK-gene-downreg-CLc
110
Epr-XOut Epr-gene-CLc
26
FosB-CLcFosB-gene-CLc
69
FtsH-CLm FtsH-gene-CLc
94
ClpC-gene-CLc
66
ClpC-CLc
65
ClpP-CLcClpP-gene-CLc
152
ComK-CLc
ComK-gene-CLc
100
Bpr-XOutBpr-gene-CLc
77
AprE-XOut AprE-gene-CLc
Abh-gene-CLc
125 127
SipS-gene-CLc SipS-CLmSipT-CLmSipT-gene-CLc
SpeB-gene-on-CLc
170
proSpeB-XOutSpeB-gene-CLc
76
137169
144
RapD-gene-CLcRapD-CLc
154
SigmaF-CLc
188
SigmaF-gene-on-CLc
192
SigmaF-gene-CLc
SigmaF-act-CLc
191
Sporulation-Sig
104
NprB-gene-CLc NprB-XOut
102
Mpr-XOutMpr-gene-CLc
RapD-gene-on-CLc
148 149
NprE-gene-CLc
98
NprE-XOut
LonB-CLc
168
HtrA-CLm
139
141
HtrA-gene-CLc
LonB-gene-CLc
12
Heme:IsdE-CLm
1615
IsdI-CLcIsdG-CLc
IsdH-CLw
3
814
IsdC-anchored-CLw
SigmaW-act-CLc RsiW-degraded-CLm
KinA-inhib-CLc
KipI-CLc
83
KinA-CLc
82
IsdD-CLm IsdF-CLm
YwlE-CLc
ClpC-inhib-CLc YpbH-CLc
UDP-GlcNAc-CLc
1
SrtA-CLm
2
MurG-CLm
IsdB-CLw
IsdC-CLw
4
SrtB-CLm IsdE-CLm
SpoIIAB-degraded-CLc
YjbH-CLc
74
SpoIIE-CLmSpoIIE-gene-CLc
187
SpoIIAB-gene-CLc SpoIIAB-CLc
185
SpoIIAA-gene-CLc
RsiW-CLm
RsbW:SigmaB-CLc
RsiW:SigmaW-CLc
92
Spo0E-gene-CLcSpo0E-CLc
Spo0E-gene-on-CLc
Spo0A-gene-CLc
KinB-CLm
85
KinC-CLc
86
78
87
KinD-CLc
PrsW-CLmSigmaF:SpoIIAB-CLc
SepF-CLc
FtsZ:SepF-CLc
RsbU-CLc
SigmaB-act-CLc
RsbT-CLc RsbW-CLc
177
RsbW:RsbV-CLc
RsbV-CLc
YlmF-CLc
FtsZ:SpoIIE-CLm FtsZ:ZapA-CLc
ZapA-CLc
FtsZ:YlmF-CLcDegU-inhib-CLc
RapD-gene-downreg-CLc
RsbS-CLc
FtsZ:FtsA-CLc
RghR-CLcRopB-CLc
pbpE-gene-downreg-CLc
FtsL-degraded-CLm
FtsL-CLm
34
RasP-CLm
FtsA-CLc
StreptolysinS-inhib-XOut
6
Heme:IsdH-anchored-CLw
StreptolysinS-XOut RapC-gene-CLc
RapF-gene-downreg-CLc
RapF-gene-CLc
RapC-gene-downreg-CLcSrfA-gene-downreg-CLc
IsdA-CLw
IlvB-degraded-CLc
IlvB-CLc
SpeB-XOut
GlmS-CLc
GlmS-degraded-CLcPurF-degraded-CLc
PurF-CLc PyrB-CLc
PyrB-degraded-CLc
RapB-CLc
5
Heme:IsdB-anchored-CLw
Heme-XOut
initial states
possible reachable states
transition/rules
consumed/produced
enzymatic relation
106
FtsH-CLm
Spo0A-gene-on-CLc
75
93
Spo0E-degraded-CLc
90
Spo0B-CLc
Spo0B-phos-CLc
Spo0E-CLc
76
Spo0A-CLc
Spo0A-gene-CLc
Vpr-XOut
Vpr-gene-CLc
Vpr-gene-on-CLc
103 97
Epr-gene-on-CLc
99
Epr-XOut
110
Epr-gene-CLc
Spo0A-phos-CLc
NprE-gene-CLcNprE-
XOut
109
NprE-gene-on-CLc
104
NprB-gene-CLc
Bpr-gene-CLc
100
Bpr-gene-on-CLc
Bpr-XOut
98
Mpr-gene-on-CLc
101105
Mpr-gene-CLc
102
NprB-gene-on-CLc
NprB-XOut Mpr-
XOut
Protease interaction network (Gram Positive Bacteria) (Anupama) What regulates a protease,
what does it regulate
What happens if you knockout FtsH?
Multiple synthesis routes
TB-mycolic-acid synthesis (Malabika)
A HypotheticaL Model PathwayRelating State and Synaptic Plasticity
Wake state: unknown signal(s) => phosphorylation of Rock1 => activation of Limk1 => phosphorylation of cofilin => increase in polymerized actin (Phosphorylated cofilin is unable to depolymerize actin)
SWS:RhoDG11 binds Rhob-GDP (is not phosphorylated) => Rock1, Limk1, and cofillin would not be phosphorylated and => actin depolymerization => decrease in synaptic weight
Challenges and Opportunities
We need good models not just good analysis tools
• Alternative computational methods to check a prediction (example: proximity)
• Type checking: using basic biological knowledge
•P has TS phosphorylation sites, Q is a tyrosine kinase
•A rule that says Q phosphorylates P is either wrong or missing intermediate steps
• Better ways to capture experimental data.
•Semantic data mining
•Symbolic supplementary data
• Inferring rules from experimental data.
•Needs representation of relevant biological knowledge
•What do experiments tell you?
•Roles -- adaptor, scaffold, transformer
Model Integration
• Aggregation -- different representation of the same process
• PID EgfR signalling + Reactome EgfR signalling
• Complementation -- different parts of a response
• Making a model executable
• Reactome pathways are disconnected -- missing rules, multitple levels of detail -- when are two entities the same?
• Mutiple computational models
• multi scale
• qualitative + quantitative
Complexity
• Query of complex models may give complex answers (all paths)
• How to visualize, make sense of the results
• Scaling
• Divide and conquer
• Property preserving abstractions
• zooming in and out
Synergy
• Finding a shared representation of relevant biological knowledge
• Combining results from multiple analyses
• how can our different tools work together?
• Biologist friendly scripting
The end! (The beginning?)