overview – systems biologysystems biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... ·...
TRANSCRIPT
![Page 1: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/1.jpg)
Overview – Systems BiologyOverview Systems Biology
![Page 2: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/2.jpg)
Trends in Scientific Investigation
20th CenturyGenomics
20th CenturyBiomedical Sciences
Reductionism
Genomics Biomedical Sciences
Genes ProteinsCell
OrganOrganismOrganism
IntegrativeComplexity
21st CenturyGenomics
ProteomicsMolecular biophysics
21st CenturyIntegrated systems biologyDynamic systems modeling
PASI 2008 Lecture Notes © Francis J. Doyle III
ComplexitySystems Analysis
o ecu a b op ys csBioinformatics
Structural Biology [adapted from (J. Weiss, 2003)]
![Page 3: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/3.jpg)
Central Dogma and Levels of Control
PASI 2008 Lecture Notes © Francis J. Doyle III
[Alberts et al., Essential Cell Biology, 1998]
![Page 4: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/4.jpg)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 5: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/5.jpg)
![Page 6: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/6.jpg)
“Engineering” of Biological Networks[Alon, 2003][ , ]
• Modularity(in network) set of nodes that have strong interactions and a– (in network) set of nodes that have strong interactions and a common function
– has defined input nodes and output nodes that control the interactions with the rest of the networkinteractions with the rest of the network
– has internal nodes that do not significantly interact with nodes outside the module
• Robustness to component tolerances
Recurring circuit elements• Recurring circuit elements
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 7: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/7.jpg)
Motifs in Biological Regulation - Yeast
PASI 2008 Lecture Notes © Francis J. Doyle III
[Lee et al., 2002]
![Page 8: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/8.jpg)
Motifs in Biological Regulation – E. coli
[Shen-Orr et al., 2002]
![Page 9: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/9.jpg)
E. Coli Transcriptional Network
[Shen-Orr et al., 2002]
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 10: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/10.jpg)
Gene RegulationGene Regulation
![Page 11: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/11.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation[Savageau, Chaos, 2001; Alberts et al., Mol. Biol. Of the Cell 4th ed., etc.]
• Transcriptional UnitsTranscriptional Units
mRNA
protein
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 12: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/12.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional UnitsStimulus
Transcriptional Units• Input Signal
mRNA
protein
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 13: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/13.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional UnitsTranscriptional Units• Input Signal
M d• ModemRNA
protein
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 14: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/14.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional Units Negative RegulationTranscriptional Units• Input Signal
M d
Negative Regulation
• ModeGene Off
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 15: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/15.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional Units Negative RegulationTranscriptional Units• Input Signal
M d
Negative Regulation
• ModemRNA
protein
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 16: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/16.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional Units Positive RegulationTranscriptional Units• Input Signal
M d
Positive Regulation
• ModemRNA
protein
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 17: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/17.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional Units Positive RegulationTranscriptional Units• Input Signal
M d
Positive Regulation
• ModeGene Off
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 18: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/18.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional UnitsTranscriptional Units• Input Signal
M d
TF1 TF2
• Mode• Logical Unit
mRNA
protein
TF1(+) TF2(-) EXP
ON ON OFFON ON OFF
ON OFF ON
OFF ON OFF
OFF OFF OFF
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 19: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/19.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional UnitsTranscriptional Units• Input Signal
M d• Mode• Logical Unit
mRNA
• Expression Cascadeprotein (TFs)
mRNA
protein (enzyme)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
M1M4M3
M2M5
metabolites
![Page 20: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/20.jpg)
Hierarchy of Biological RegulationHierarchy of Biological Regulation
• Transcriptional UnitsTranscriptional Units• Input Signal
M d• Mode• Logical Unit mRNA
• Expression Cascade• ConnectivityConnectivity
mRNA
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 21: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/21.jpg)
Gene Regulatory NetworksGene Regulatory Networks
• Whole genome sequencing (all) potential g q g ( ) pmacromolecular players
• High throughput methods are at relatively mature state
GRN models DNA specific predictions (which can be• GRN models DNA-specific predictions (which can be validated)
• Transcription and translation are slow (cf. protein-protein and enzymatic rxns). May suggest switch-like behavior (B l ) f l tt
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
(Boolean) for latter.
![Page 22: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/22.jpg)
Eukaryotes – More Complex StoryEukaryotes More Complex Story
• Gene regulatory proteins can influence from a long distance (thousands of bp away from promoter) single promoter influenced by virtually unlimited number of regulatory sequences scattered along DNA
• RNA polymerase II (transcribes all protein-coding genes) cannot initiate transcription alone. Requires “general” transcription factors to be assembled.
• Packing of eukaryote DNA into chromatin provides additional layers g y p yof regulation (not available to bacteria)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 23: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/23.jpg)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 24: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/24.jpg)
Integrated Circuits
• Multi-gene interactions
• Already studied: genes are regulated
• Will study: proteome is dynamic – changing w/ environment– But promoters don’t change…– How do cells turn “on” and “off”?…– [and not in a wildly fluctuating manner…]
• One interesting motif: toggle switch (turn genes on and off)– Ramifications for development, cell cycle, cancer, etc.
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
– Interesting attribute: robust probabilistic switching under uncertainty
![Page 25: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/25.jpg)
Noise & Uncertainty
• Cell division/mitosis & cytokenesis how to divide TFs– “randomness” captured w/ binomial frequency functionrandomness captured w/ binomial frequency function– e.g., 50 TF, 2 daughters, p=0.50– 50/50 split 0.112
25 / 5 0 880 t
knk ppkn −−⎟⎟
⎠
⎞⎜⎜⎝
⎛)1(
– 25+/-5 0.880, etc.
• Few binding sites for each protein, slow rates of binding“ ” / /
k ⎟⎠
⎜⎝
– “randomness” right protein/time/place– also, once bound, variable delay for activation of transcription– protein produced from one gene obeys normal distribution(i.e., random variable that is composite of many small random
events)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
• Environmental uncertainty– development, morphology, concentrations, etc.
![Page 26: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/26.jpg)
Toggle Switch Circuit
Protein A: TF for genes b & c
Protein B: degraded by cell, performs other functions, represses expression of c
Protein C: degraded by cell, perform other functions, represses expression of bg y , p , p p
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 27: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/27.jpg)
Toggle Switch Circuit
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
[Campbell & Heyer, 2006]
![Page 28: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/28.jpg)
λ Phage Switch – Introduction[Arkin et al., Genetics, 1998]
• Pathogen (virus) that can switch exterior to fool immune system
• Two life domains– Live quietly in E. coli (lysogenic)Live quietly in E. coli (lysogenic)– Replicate quickly, kill host, launch progeny (lytic)
• Choice determined by single protein: CII• Choice determined by single protein: CII
• Very similar to simple circuit just considered
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 29: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/29.jpg)
λ Phage Switch Circuit
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
[Campbell & Heyer, 2006]
![Page 30: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/30.jpg)
How do Cells Cope?
• Usual control tricks
– Cascades and relays (low pass filters)
– Negative feedback
– Integral feedbackg
– Redundancy
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 31: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/31.jpg)
Modeling IssuesModeling Issues
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 32: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/32.jpg)
Iterations for Model and Hypothesis
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
[Kitano, Nature, 2002]
![Page 33: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/33.jpg)
The Spectrum of Descriptions[Stelling, 2004]
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 34: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/34.jpg)
Nice Example of Modeling Scope
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
[Bolouri & Davidson, 2002]
![Page 35: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/35.jpg)
Nonlinear ODE Representations
• Gene regulation captured by rate equations: nixfdtdx
ii ≤≤= 1)(
• Can also incorporate discrete time delays:
dt
dx
E l k d d t f th i hibit i f
nitxtxfdtdx
inniii ≤≤−−= 1))(,),(( 11 ττ K
• Early work: end product of pathway co-inhibits expression of gene coding for enzyme that catalyzes step in pathway
22122
11311 )(
xxkxxxkx
xxrkx
γγ
γ
−=−=
−=
&
&
& mRNA
A
K
MA
AKR
a
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
33233 xxkx γ= K
FKR
![Page 36: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/36.jpg)
Enzymatic Kinetics
Michaelis-Menten Kinetics
PEESSEkk
k+→⇔+
−
21
1
Assumptions:1) [S] >> [E]1) [S] [E]2) Steady-state ([ES] ~ constant)3) Total enzyme concentration is constant
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 37: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/37.jpg)
A2 + algebra: A3 + algebra:
[ ][ ][ ][ ] [ ]ESK
SE
M
= [ ] [ ] [ ]
[ ][ ]
ESEE +=0
1
21
kkkKM
+−=
[ ][ ][ ]SKSEkv
M += 02
1
[ ]Svv = max
[ ]SKv
M +=
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 38: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/38.jpg)
Enzymatic Kinetics
Hill Kinetics( b t t l l bi d t )(n substrate molecules bind to enzyme)
PCSE →⇔+ 1
PCCCSPCCCS
+→⇔+
+→⇔+
23`2
12`1 [ ][ ]nn
M
n
SKSvv
+= max
PCCCS nnn +→⇔+ −− 1`1
M[ ]M
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 39: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/39.jpg)
Transcriptional Regulation Module(Adapted from [Barkai and Leibler 2000])(Adapted from [Barkai and Leibler, 2000])
jk
PikPij Mi i i2+ -j2k2
Pmi
Pli +
+
2 22 2[ ] [ ] [ ] [ ] [ ] [ ]RPij Rj Pij RPik Rk Pik dMjj kMi Mid k k k k kdt
Pij Pij Pik Pik= + + + −
2 22 2
2
[ ] [ ][ ] [ ]
[ ] [ ][ ] [ ]
pij upijj jd k kdt
d k k
Pij Pij Pij
Pij Pij Pijj j j
= − +
2 22
22
2 2
22
[ ] [ ][ ] [ ]
[ ] [ ] [ ] 2 [ ] 2 [ ]
pij upij
Ti di i uiMi
k kdt
d i k k i k
j Pij Pijj j j
i kdt
i
= −
= − − +
22
2[ ] [ ]i
dtd ki
ti
d= 2 22 2[ ] [ ] { } { }ui dik k promoter binding promoter unbindi ingi− − − +
![Page 40: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/40.jpg)
Comments
• Overall transcription = linear sum of bound and unbound promoter sites
• In example:p– TF j2 increases rate kRj2Pij >> kRPij
– TF k2 represses i kRPik >> kRk2Pik
– If binding k2 blocks activation j2 kRj2Pij >> k (both bound)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 41: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/41.jpg)
Pros/Cons of Nonlinear ODE Models
• Not analytically solvable• Host of powerful integration engines• Time delay adds challenge (especially variable time delay)• Often get lots of parametersOften get lots of parameters
• Challenge: combinatorial explosion in parametersO t it id tif iti it• Opportunity: identify sensitivity
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 42: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/42.jpg)
Computational Models of Chemical Reacting Systemsg y
• Continuous and deterministic – (rate equations) Described ( q )by ordinary differential equations (ODE). Huge numbers of molecules.
• Continuous and stochastic – (Langevin regime) Valid under certain conditions. Described by Stochastic Differential Equations (SDE) Large n mbers of molec les(SDE). Large numbers of molecules.
• Discrete and stochastic – Finest scale of representation forDiscrete and stochastic Finest scale of representation for well stirred molecules. Exact description via Stochastic Simulation Algorithm (SSA) [Gillespie, 1977]. The only algorithm for small numbers of molecules.
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 43: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/43.jpg)
Continuum Versus Stochastic Simulation
ContinuumStochastic
• Molar (mole/L)• # of molecules
P it f ti( )
• Reaction rates• Propensity functions
(probability of rxns)• ODEs describing the
• ODEs describing the changes in states
• ODEs describing the changes in probabilities of states (Master equation)
(reactants, products)• Initial value problem
solver
• Stochastic simulation algorithm (SSA)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
solver
![Page 44: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/44.jpg)
4 Possible States
J2k01
0 1
Pij Pik Pij Pik
2
k10
k01
k20 k31k02 k13
2 3
Pij PikPij Pik
J2K2 K2k232
k32
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 45: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/45.jpg)
“Conservation” Equations
( ) ( )k →01
( )( ) ( )( )
k
ikijkikij PPJPPJ •⎯⎯←⎯→⎯
+ 22 11110
( )( ) ( )( )ijikk
k
ikij PPKPPK •⎯⎯←⎯→⎯
+ 22 11120
02
( )( ) ( )( )ikijk
k
ijik PKPJPPKJ ••⎯⎯←⎯→⎯
•+ 2222 11132
23
( )( ) ( )( )ikijk
k
ijik PKPJPPJK ••⎯⎯←⎯→⎯
•+ 2222 11131
13
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 46: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/46.jpg)
How to Simulate
• Monte Carlo methods to simulate molecule state changes
• Alternative: capture probabilities directly– Molecules represent configuration– State is probability distribution over all configurationsState is probability distribution over all configurations– # probabilities scales with # configurations– Obeys differential equation
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 47: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/47.jpg)
Simple Example
),0( ttP⎥⎤
⎢⎡ Δ+
)(),2(),1(
)( tPAttPttP
ttP =
⎥⎥⎥⎥
⎢⎢⎢⎢
Δ+Δ+
=Δ+
),3( ttP ⎥⎦
⎢⎣ Δ+
)(][10][
0][1][0][][1
3222320202
3121310201
2010202201
tPtktJktktKktktKktktJk
tktktKktJk
⎥⎥⎥⎥⎤
⎢⎢⎢⎢⎡
ΔΔ−Δ−ΔΔΔ−Δ−Δ
ΔΔΔ−Δ−
=
1][][0][][
3231223213
3222320202
tktktJktKk ⎥⎦
⎢⎣ Δ−Δ−ΔΔ
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 48: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/48.jpg)
ODE Version
• Take limit of small time step )(tPBPd= )(tPB
dt=
– No such thing as equilibrium configuration– There is an equilibrium in probability distribution
M lti l l ith il bl t l– Multiple algorithms available to solve
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 49: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/49.jpg)
Chemical Master Equation (CME)
• Key assumption: well-stirred system
• Reactions are discrete random events with probability given by the propensity function
aj(x)dt : the probability, given X(t) = x, that one Rj reaction happen insides Ω in between the time interval [t,t+dt)
• No exact prediction of states, but can track the probabilityP(x,t|x0,t0) : the probability of X(t) = x, given the initial condition X(t0) = x0 (t>t0)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 50: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/50.jpg)
Chemical Master Equation Derivation
( )+ =0 0, | ,P t dt tx x ( ) ( )⎡ ⎤−⎢ ⎥∑0 0, | , 1
M
jP t t a dtx x xthe system is already in state xd ti
( )0 0
( ) ( )− −∑ 0 0, | ,M
j j jP t t a dtx x xν ν
( ) ( )=
⎢ ⎥⎣ ⎦
∑0 01
, | , jjand no reaction occurs
the system is in state x–νj and + ( ) ( )=
∑ 0 01
, | , j j jj
P t t a dtx x xν νjreaction Rj occurs
+
( ) ( ) ( ) ( ) ( )∂⎡ ⎤∑0 0, | ,
| |MP t tx x
Chemical Master Equation
( ) ( ) ( ) ( ) ( )=
∂⎡ ⎤= − − −⎣ ⎦∂ ∑0 0
0 0 0 01
, | ,, | , , | ,j j j j
j
t ta P t t a P t t
tx x x x x xν ν
… in general, there exists no analytical solution
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
g y
![Page 51: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/51.jpg)
Stochastic Simulation Algorithm (SSA)[Gillespie, 1976]
• Main idea:
p(τ,j|x,t)dτ : the probability, given X(t) = x, that the next reaction will occur in within [t+τ,t+τ+dτ), and will be an Rj reaction
• Joint probability function of “time to next reaction” (τ )and “index of next reaction” ( j )and index of next reaction ( j )
p(τ,j|x,t)dτ is a function of the propensities
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 52: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/52.jpg)
Stochastic Simulation Algorithm
1. Initialize the time t = t0 and the state x = x0
2 Evaluate the propensities a (x)2. Evaluate the propensities aj(x)3. Pick two random numbers from uniform distribution4 Compute τ and j4. Compute τ and j
( )τ
⎛ ⎞= ⎜ ⎟
⎝ ⎠∑ 1
1 1lnia rx( ) ⎝ ⎠∑ 1ia rx
( ) ( )=
= ≥∑ ∑21
smallest integer : j
i ii
j a r ax x
5. Step forward in time by τ and update the states x(t+τ) = x+νj
1i
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
6. Repeat from beginning
![Page 53: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/53.jpg)
Stochastic Gene Network Response
5
1
2
GC [m
olec
ules
]
0 )
3
4
5
[nM
] MD
MF MB
Q035
0 10 20 30 40 50 60 70 80
0P
15
20
ecul
es]log(
Mi/M
i0
0
1
2
Liga
nd in
put
MA MH
MG & MK MJ
0 10 20 30 40 50 60 70 800
5
10
Time [h]
MG
[mol
e
Time [h]0 10 20 30 40 50 60 70 80
-2
-1
L
MC
ME MA
![Page 54: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/54.jpg)
Stochastic Gene Network Response
60 103
40
50
ules
]
102
10
les2 ] MB
20
30
MA
[mol
ecu
101
σ2 M
i [mol
ecul
MA
0 10 20 30 40 50 60 70 800
10
Time [h]10 20 30 40 50 60 70 80
100
Time [h]
MC
![Page 55: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/55.jpg)
500
1000Q 500
1000
Q
mRNAs (molecules) vs. time (hours)
0 20 40 60 800
500Q
0 20 40 60 800
500Q
100
200
MA 200
400
MB
0 20 40 60 800
100M
0 20 40 60 800
200M
100
200
MC 50
100
MD
0 20 40 60 800
M
0 20 40 60 800
M
20
40
ME 100
200
MF
0 20 40 60 800
M
0 20 40 60 800
M2
4
MG 60
80
MH
0 20 40 60 800
M
0 20 40 60 8040
M
50
100
MJ 1
2
MK
0 20 40 60 800
hours0 20 40 60 80
0
hours
![Page 56: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/56.jpg)
500
1000
Q 500
1000
Q
Transcription factors (molecules) vs. time (hours)
0 20 40 60 800
500Q
0 20 40 60 800
500Q
3x 104
10x 104
0 20 40 60 800
1
2
A
0 20 40 60 800
5CC
6x 104
6x 104
0 20 40 60 800
2
4
DD
0 20 40 60 800
2
4
FF
0 0 0 60 80 0 0 0 60 80
0
100
200
300
GG
0
100
200
300
KK
0 20 40 60 800
0 20 40 60 800
hours
5
10x 104
EQ
0 20 40 60 800
hours
![Page 57: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/57.jpg)
Observations on Stochastic Simulation
• One can go even deeper– Molecular dynamicsy– Quantum mechanical description
• But not tractable for gene regulation (nor insightful)But not tractable for gene regulation (nor insightful)
• Guidelines:
1. # molecules sufficiently high that single molecule changes can be approximated by change in continuous concentration ODE
2. Fluctuations about mean << mean ODE
3. Otherwise, and if solution well mixed locally stochastic
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 58: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/58.jpg)
Lambda PhageLambda Phage
![Page 59: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/59.jpg)
Lambda Phage Revisited
• Virus that infects E. Coli• 2 developmental pathways
– Replicate/lyse– Lysogeny
• Simple developmental switch• 50 000 bp genome sequenced early on lots of regulatory knowledge50,000 bp genome sequenced early on, lots of regulatory knowledge• Architecture suggests two approaches to analysis:
– Boolean circuits/switches– Stochastic distribution – bistable switch
• References:– [Bower & Bolouri Ch 2][Bower & Bolouri, Ch. 2]– [McAdams & Shapiro, 1995]– [Arkin et al., 1998]
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 60: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/60.jpg)
Lambda Phage “Circuit”
PASI 2008 Lecture Notes © Francis J. Doyle III
[Campbell & Heyer, 2006]
![Page 61: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/61.jpg)
Discrete Stochastic Model[Gibson/Bruck]
• Focus on N, Cro2, and CI2– CI2 is present at high levels in lysogens and represses expression of all 2 p g y g p p
other genes– Cro is key in lytic pathway: inhibits production of CI2 and controls
production of key proteins (cell lysis, replication of λ DNA)– CI2 and Cro are mutually inhibitory– N produced early in life cycle, production halted after fate choice
• Model: Regulation of N as function (Cro and CI2)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 62: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/62.jpg)
Final time distribution of cro2 and repressor
PASI 2008 Lecture Notes © Francis J. Doyle III
[Gibson & Bruck, Caltech Tech Report 26]
![Page 63: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/63.jpg)
Basic Model for N Production
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 64: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/64.jpg)
Basic Model – TF-DNA Binding
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 65: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/65.jpg)
Dependence of Transcription Initiation
PASI 2008 Lecture Notes © Francis J. Doyle III
[Bower & Bolouri, 2002]
![Page 66: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/66.jpg)
Probability of N at time t
PASI 2008 Lecture Notes © Francis J. Doyle III
[Bower & Bolouri, 2002]
![Page 67: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/67.jpg)
Impact of Repressor
PASI 2008 Lecture Notes © Francis J. Doyle III
[Bower & Bolouri, 2002]
![Page 68: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/68.jpg)
Questions to Address
• What is probability that the given mRNA will start translation rather than be degraded & vice versa?
• What is the probability that n proteins will be produced from one mRNA molecule before it degrades?g
• What is the average number of proteins produced/mRNA transcript?
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 69: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/69.jpg)
Detailed Circuit Schematic[Arkin et al., Genetics, 1998]
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 70: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/70.jpg)
Key Assumptions
1. Cell generation time is deterministic
2 Linear growth in volume2. Linear growth in volume
3. Housekeeping molecules constitutively expressed
4. .
5. .
6. .
7 Gene expression is stochastic7. Gene expression is stochastic
8. .
9. .
10. Target cells are infected simultaneously
11. Well mixed (cell)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 71: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/71.jpg)
Detailed ModelTranscription/Translation
![Page 72: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/72.jpg)
Detailed ModelHousekeeping/Nongenetic Elements
![Page 73: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/73.jpg)
![Page 74: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/74.jpg)
![Page 75: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/75.jpg)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 76: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/76.jpg)
Insights Gained from Systems Biology Approachy gy pp
• Study reveals how thermal fluctuations can be exploited by the regulatory circuit designs of developmental switches to produce different phenotypic outcomes
S ifi l i b t l f t i ti it l l f• Specific conclusions about role of termination sites on level of lysogeny (in silico mutations)
• Generic switch insights
• Robust yet random performance (hypothesis: dispersion in timing across population and not dispersion in outcome)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 77: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/77.jpg)
Circadian RhythmCircadian Rhythm
![Page 78: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/78.jpg)
Circadian Rhythms
Circadian rhythms = self-sustained biological rhythms characterized by a free-running period of about 24h (circa diem)
Circadian rhythms characteristics:• General – bacteria, fungi, plants, flies, fish, mice, humans, etc., g , p , , , , ,• Entrainment by light-dark cycles (zeitgeber)• Phase shifting by light pulses• Temperature compensation
Circadian rhythms occur at the molecular level
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
y
![Page 79: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/79.jpg)
Circadian Rhythm
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
Source: The Body Clock Guide to Better Health, 2000
![Page 80: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/80.jpg)
Circadian Rhythm and Gene Studies[M. Rosbash, HHMI]
![Page 81: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/81.jpg)
Connections to Sleep Disorder
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
[Wagner-Smith & Kay, Nat. Genet., 2000]
![Page 82: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/82.jpg)
Circadian Gene Regulation• Cellular circadian rhythmicity arises from a complex
transcriptional feedback structure
• Several model systems have generated insight– Drosophila
Neurospora– Neurospora– Mouse
• Tremendously robust regulatory architecture• Tremendously robust regulatory architecture
• Key structural elementsy– Autoregulatory transcriptional/translational negative feedback loop(s)– Positive feedback loop(s) between autoregulatory loops
(clock, period/time)P t i i ti d l ( h h l ti di i ti t t)
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
– Protein processing time delays (phosphorylation, dimerization, transport)
![Page 83: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/83.jpg)
Drosophila Circadian Oscillator
PERPER
PDBT
PER
PERP
TIM
PERTIM TIM
P
TIM
TIM
DBT
TIMPper
tim
TIM
Cytoplasm
Nucleus
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
y p
![Page 84: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/84.jpg)
Circadian Rhythm Gene Network[Drosophila]
pertim dClk
dCLKCYC
VRI PDP1ε
PERTIM
vriPdp1ε
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
adapted from [Cryan et al., Cell, 2003]
![Page 85: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/85.jpg)
Circadian Rhythm Gene Network[Mouse]
percry Clk
Key numbers:CLK
Bmal1REV
ERBαRORα
Key numbers:~60-100 mRNA~1000-1500 protein~10,000 neurons in SCN
PERCRY
rev erbαrorα
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
adapted from [Cryan et al., Cell, 2003]
![Page 86: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/86.jpg)
Generic Model[Forger et al., 2003]
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 87: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/87.jpg)
Model Validation Criteria[Forger et al., 2003]
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 88: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/88.jpg)
5-State Model [Goldbeter, 1996; Gonze et al., 2002]
5 t t 18 tn
P I Ps mn n
I N m P
dM K Mdt K P K M
ν ν= −+ +
5 states, 18 parameters
0 0 11 2
1 0 2 1
01 1 1 2
I N m P
s PdP P Pk Mdt K P K P
PdP P P P
ν ν= − ++ +
01 1 1 21 2 3 4
1 0 2 1 3 1 4 2
2 1 2 23 4 1 2 2
3 1 4 2 2d N
d
dt K P K P K P K P
dP P P P k P k Pdt K P K P K P
ν ν ν ν
ν ν ν
= − − ++ + + +
= − − − ++ + +
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
3 1 4 2 2
21 2 2
d
NdP k P k Pdt
= −
![Page 89: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/89.jpg)
10-state Model [Gonze et al., 2002]
10 states, 38 parameters
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 90: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/90.jpg)
Entrainment Behavior[Leloup & Goldbeter]
![Page 91: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/91.jpg)
Light as a Zeitgeber
2
4
6(h
)Advance
-2
0
2
ha
se
sh
ift
(
- 6
-4
2
0 4 8 1 2 16 20 24
P
Delay
0 4 8 1 2 16 20 24
Initial phase (h)
Experimental data [Hall & Rosbash 1987]
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
Experimental data [Hall & Rosbash, 1987]
Theoretical phase response curve [LeLoup]
![Page 92: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/92.jpg)
Influence of Light Pulses [Bagheri et al., 2004]
4
01234
tion
50 60 70 80 90 100 110 120 130 140 1500
234
Con
cent
ra
50 60 70 80 90 100 110 120 130 140 15001
Prot
ein
C
34
T i h
P
50 60 70 80 90 100 110 120 130 140 150012
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
Time, hours
![Page 93: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/93.jpg)
Stochastic Model - Simple
0 0 11 2
1 0 2 1
nP I P
s mn nI N m P
s P
dM K Mdt K P K M
dP P Pk Mdt K P K P
ν ν
ν ν
= −+ +
= − ++ +
01 1 1 21 2 3 4
1 0 2 1 3 1 4 2
2 1 2 23 4 1 2 2
3 1 4 2 2
21 2 2
d Nd
N
PdP P P Pdt K P K P K P K P
dP P P P k P k Pdt K P K P K P
dP k P k Pd
ν ν ν ν
ν ν ν
= − − ++ + + +
= − − − ++ + +
= −1 2 2 Ndt
![Page 94: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/94.jpg)
Stochastic Model - Detailed
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 95: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/95.jpg)
![Page 96: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/96.jpg)
Enzymatic Degradation of mRNA
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 97: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/97.jpg)
Stochastic Behavior[Ω=100]
Deterministic Simple Stoch Complex Stoch
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 98: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/98.jpg)
Stochastic Behavior[Ω=50]
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 99: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/99.jpg)
Implications from Systems Biology Studies
• Robustness characteristics of feedback architecture d t h ti t i tunder stochastic uncertainty
• Underlying design principles
• Nature of entrainment, and systems characterization
• Possible therapeutic ramifications (mutants, etc.)
G l bi l i l ill t i i ht• General biological oscillator insights
ACC Short Course Lecture Notes © Francis J. Doyle III, June 2006
![Page 100: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/100.jpg)
Biomedical ControlBiomedical Control
![Page 101: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/101.jpg)
The Glucose – Insulin “Loop”(i) Automatic Control
?(i) Automatic Control(ii) Day-to-day Control(iii) Efficient Solution?
Glucose Insulin
PASI 2008 Lecture Notes © Francis J. Doyle III
MeasurementInsulin
Delivery
![Page 102: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/102.jpg)
Episodic Measurements:Run-to-Run Control
![Page 103: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/103.jpg)
Run-to-Run ControlPreliminaries
• Iterative Learning Control (ILC)– arose from repetitively operated systemsarose from repetitively operated systems
• antenna servomechanism– also useful for switching between inputs
• robot actions• robot actions
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 104: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/104.jpg)
Learning Control Scheme
B i l ith( ) ( )k d ke y t y t= −
• Basic algorithm:
• For LTI plant ([A B C D]) convergence:
1k k ku u e+ = + Γ &
1i
I CB− Γ <For LTI plant ([A,B,C,D]), convergence:lim ( ) ( )
i
k dky t y t
→∞→
• Algorithm generalization:
u u e e e dt= + Φ + Γ + Ψ∫&1k k k k ku u e e e dt+ = + Φ + Γ + Ψ∫
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 105: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/105.jpg)
Optimization-based r2rPreliminaries
• Optimization-based r2r– model-based frameworkmodel-based framework
• gradient-based update between iterations– model-free framework
• terminal constraint handling• terminal constraint handling
dController
Update Law SSd
n
π ΨΨr
n
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 106: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/106.jpg)
Optimization Problems
• Nominal Problem:( )
min φ( ( ))fu tJ x t=
0s.t. ( , ); (0) ( , ) 0, T( ( )) 0f
x f x u x xS x u x t
= =≤ ≤
&
• Robust Optimization:
( )min φ( ( ))fu t
J x t=( )
0s.t. ( ,θ, ) ; (0) ( , ) 0, T( ( )) 0
u t
f
x F x u d x xS x u x t
= + =≤ ≤
&
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 107: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/107.jpg)
Optimization Problems (cont’d)
• Measurement-based Optimization (MBO):
( )min φ( ( ))
k
k kf
u tJ x t=
0
s.t. ( ,θ) ( ,θ) (0)
k k k k k
k
x f x g x u dx x
= + +
=
&
( ,θ) ( , ) 0
( ( )) 0
k k k
k k
k
y h x vS x u
= +
≤
T( ( )) 0
given ( ) , ( 1)
kf
j
x t
y i i j k
≤
∀ ∀ ≤ −
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 108: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/108.jpg)
Parameterization of Input Vector
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 109: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/109.jpg)
Classification of Measurement-based Optimization Approachesp pp
• Fixed model MBO (repeated)• Fixed model – MBO (repeated)• accuracy of model
• Refined mode – MBO (repeated)• persistency of excitation
• Evolutionary Approaches• curse of dimensionality• curse of dimensionality
• Reference Tracking• what to track for optimality
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 110: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/110.jpg)
Application SummaryApplication Summary
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 111: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/111.jpg)
Simulated Chemical Reactor
• (Srinivasan et al., 2001)
• Optimize productivity of complex reaction
• Approach:• Feed rate is actuator, parameterized by 3-D
N d l i l i t i• No model – simple gain matrix• Optimization-based approach• Complex constraints
• Effective convergence in ~10-20 batches
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 112: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/112.jpg)
Algorithm Architecture
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 113: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/113.jpg)
Results
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 114: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/114.jpg)
Experimental Chemical Reactor
• Lee & Lee in (Bien & Xu, 1998)
• Reactor temperature recipe (5000 sec) • (charge,heat-up,reaction,cooling,discharge)
• Approach:J k t t t fil i t t• Jacket temperature profile is actuator
• Inaccurate ARX model employed• Combined iterative and feedback algorithm
• Effective convergence in ~6 batches
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 115: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/115.jpg)
Functional Neuromuscular Stimulation (FNS)( )
• Dou et al. in (Bien & Xu, 1998)
• FNS of limb no longer under voluntary control
• Challenges• customization to individual patientscustomization to individual patients• adaptation from time-varying musculoskeletal system • robustness against exogenous disturbances
• Approach:• Simplified fundamental model employed
PASI 2008 Lecture Notes © Francis J. Doyle III
p p y• Track joint angle using pulse width (flexor, extensor)
![Page 116: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/116.jpg)
Algorithm Architecture
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 117: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/117.jpg)
Results
• ILC effective with muscle fatigue• ILC rejects repeated uncertainty and disturbancej p y• ILC tracks slowly varying desired trajectory
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 118: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/118.jpg)
Parameter Identification (ILI)
• (Chen & Wen, 1999)
• Aerodynamic drag coefficient modeling
• Approach:• ILC: given a reference trajectory and repeated data, refine the input
profileprofile• ILI: given I/O data (reference) and repeated data, refine uncertain
coefficient
• Effective convergence in ~10-30 cycles
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 119: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/119.jpg)
Robot Repeated Task
• (Moore, 1993)
• Two joint manipulator
• Approach:• simplified (nonlinear) fundamental model
t i th t t• torque is the actuator• adaptive gain adjustment• P-ILC algorithm
• Effective convergence in ~8 cycles
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 120: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/120.jpg)
Insulin Injection Optimization
• (Doyle III et al., IEEE EMBS Conf., 2001)
• Timing/amount of (repeated) insulin injections
• Approach:• Gain model (implicit)• Gmax and Gmin are objectivesGmax and Gmin are objectives• Fixed (decoupled) PI control structure• MBO framework
• Effective convergence in ~6-10 cycles
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 121: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/121.jpg)
Observations – Current Patient Protocol
• Availability of periodic glucose measurement
• Issues with accurate models for individual patients• “Batch-like” = single meal or 24 hour cycle• Few key variables
– input: timing and size of insulin bolus– performance: maximum and minimum glucose values
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 122: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/122.jpg)
Run-to-Run Algorithm
max max( 1) ( ) min(0, ( ))rTT k T k K G G k+ = + −
min min( 1) ( ) max(0, ( ))rQQ k Q k K G G k+ = + −
• Initial guesses T(1), Q(1)
R f l f G d G• Reference values for Gmax and Gmin
• Can impose hard bounds on Gmax and Gmin
• Gains KT and KQ reflect compromise between speed and accuracyGains KT and KQ reflect compromise between speed and accuracy
• Straightforward generalization to 3-meal (24 hr) cycle
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 123: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/123.jpg)
Open-loop glucoseDay 1Day 1
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 124: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/124.jpg)
Day 1Open-loop glucose Day 1
Day 2
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 125: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/125.jpg)
Open-loop glucoseDay 1Day 1
Day 2Day 10
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 126: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/126.jpg)
Preliminary Clinical Evaluation
• Summer 2003 @ Sansum Diabetes Research Institute@
• 9 type I patients, pump users
• Separate phases for:p p– Bolus determination– Patient sensitivity identification
Si l l t t di– Single meal run-to-run studies– 3-meal run-to-run studies
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 127: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/127.jpg)
Phase 6 (3 meal)
Corrective Action Necessary
Target = 150 mg/dl
//////////////////////////////150 mg/dl///////////////////////////// 150 mg/dl
No Action Necessary
75 mg/dl
G60-TimeG90-AmountController Pairing:
g
//////////////////////////////60 mg/dl//////////////////////////////
Corrective Action Necessary
PASI 2008 Lecture Notes © Francis J. Doyle III
Target = 75 mg/dl
![Page 128: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/128.jpg)
Preliminary Clinical Evaluation
Glucose (60 min) Glucose Bounds
Insulin Amount
Bounds
Start of
Day
Start ofalgorithm
Glucose (90 min)
Insulin TimingStart ofalgorithm
PASI 2008 Lecture Notes © Francis J. Doyle III
Day
![Page 129: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/129.jpg)
Phase 6 Results• Class A (convergent, 3-4 days, clinically adequate range) [41%]• Class B (always within range) [26%]• Class C (divergent, incorrect sensitivity, mitigating circumstances) [33%]
Patient ID Dinner Breakfast Lunch
B B BDLL02 B B B
SL11 C C A
LGW06 B A BLGW06 B A B
HDC09 A C C
JES08 A B A
LJ01 C A B
JLV07 A A A
PASI 2008 Lecture Notes © Francis J. Doyle III
MHS10 A C C
RPE05 C C A
![Page 130: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/130.jpg)
Summary - Observations
• By providing prepared meals to the patients, food chaos & variability was minimized.
• Due to the design of the trial patients needed to record 3 BG• Due to the design of the trial, patients needed to record 3 BG measurements for each meal (9/day). If site problems or other events occurred, patients often checked more frequently as this was encouragedencouraged.
• In Phase V and Phase VI, the patients were under dosed by ~25% and were still able to keep G60 and G90 between 60 and 150 mg/dl suggesting that once in good control it is difficult to be bumped outsuggesting that once in good control, it is difficult to be bumped out of control.
• As the trial progressed, the A1C values decreased significantly• The impact of the bolus timing was unclear in terms of the effect on• The impact of the bolus timing was unclear in terms of the effect on
the glucose profile.
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 131: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/131.jpg)
Comparison of Performance Metrics
Max/Min GlucoseFixed Time GlucoseGlucose Difference
Lunch start
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 132: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/132.jpg)
New Algorithm Formulation
insulin meal boluspostprandial glucose difference
• Only changing insulin dose, timing always fixed to the beginning of the meal
Still i t t l t• Still require two post-meal measurements– First measurement 60-90 minutes after the start of the meal– Second measurement 30-60 minutes after the first– For each meal, denote these times as:
T T T T T T
PASI 2008 Lecture Notes © Francis J. Doyle III
TB1, TB2, TL1, TL2, TD1, TD2
![Page 133: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/133.jpg)
Robustness of Algorithm
• variable meal timei bl b h d t t t• variable carbohydrate content
• variable measurement time• noise on measurement• incorrect meal estimate
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 134: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/134.jpg)
Clinical Evaluation of New Algorithm
• 11 subjects with type 1 diabetes & CSII pumps
• Phase 1– Optimized basal rates– Brought out of control (1h post-prandial 170–200 mg/dl)– Lunch only– Carbohydrate content kept constant– Algorithm adjusted dosing over 2 weeks
• Phase 2– All three meals– Carbohydrate content varied– Algorithm adjusted dosing over 2–3 weeks
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 135: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/135.jpg)
Challenge in Data Clustering
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 136: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/136.jpg)
Medically-Inspired Performance Measure
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 137: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/137.jpg)
Phase 1 Results
• Time to convergence: 5.4±3.6 daysO th fi t d• On the first day– pre-prandial BG: 101.7±22.4 mg/dl– 60 min post-prandial BG: 176.5±41.6 mg/dl
IC ti 1U t 14 15 3 95 b h d t– IC ratio: 1U to 14.15±3.95 g carbohydrate• On convergence
– pre-prandial BG: 94.7±23.9 mg/dl– 60 min post-prandial BG: 109.5±25.3 mg/dl– IC ratio: 1U to 9.47±2.27 g carbohydrate
• Over 118 meals, only two hypoglycemic (<55 mg/dl) events were reported (a 1.7% incidence rate), & none below 50 mg/dl
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 138: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/138.jpg)
Modifications for Phase 2
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 139: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/139.jpg)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 140: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/140.jpg)
Closing the LoopClosing the Loop
![Page 141: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/141.jpg)
Model-Based Control Approach[Parker, Doyle III, Peppas, IEEE Trans. Biomed. Eng., 1999]
Model-basedAlgorithmD i d
Controller Patient
AlgorithmDesiredGlucose Level GlucoseInsulin
-
ModelKalman Filter -
UpdateFilter
Compartmental Model
Key tenet of Robust Control Theory:
PASI 2008 Lecture Notes © Francis J. Doyle III
Key tenet of Robust Control Theory:Model accuracy is directly tied to achievable performance
![Page 142: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/142.jpg)
Moving Horizon Concept of MPC
kpast future
ktarget glucose value
predicted glucose trend
glucose measurements
k+pk+m ......k+1
projected insulin delivery
PASI 2008 Lecture Notes © Francis J. Doyle III
k+pk+m ......k+1
time
![Page 143: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/143.jpg)
MPC Components
• Reference Trajectory Specification
• Process Output Prediction (using Model)
• Control Action Sequence Computation( i bl )(programming problem)
• Error Prediction Update (feedback)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 144: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/144.jpg)
Unconstrained MPC
• Recall: )1()1()( −Δ+−= kvSkYMkY S
• Model Prediction:
)1()1|1()1|( −Δ+−−=− kvSkkYMkkY S
• Correction:
State estimate at time k-1
( ))1|()(ˆ)1|()|( −−+−= kkykykkYkkY Iestimatemeasurement
nI
I
⎪
⎪⎬
⎫
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡= MI
PASI 2008 Lecture Notes © Francis J. Doyle III
I ⎪⎭⎥⎦⎢⎣
![Page 145: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/145.jpg)
Control Problem22
1
)()|(∑=
+ΔΔ+−+
p
krkkyl
Kll
-1)mu(k,u(k),min :Objective
• Suppose we want to control some outputs more tightly than others?
( ) ⎟⎠⎞⎜
⎝⎛ =• 2
1xxT
• Suppose we want to control some outputs more tightly than others?
– premultiply by: ylΓ ll ∀⎥
⎦
⎤⎢⎣
⎡=Γ :examplefor
2
1
00
γγy
• Suppose we want to penalize manipulated variable moves?
⎦⎣ 20 γ
– premultiply by: ulΓ
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 146: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/146.jpg)
[ ] [ ]22
)1()()|( ∑∑ +ΔΓ+++Γm
up
y kukrkky lllmin [ ] [ ]11
)1()()|( ∑∑==
+ΔΔ−+ΔΓ++−+Γ y kukrkky
ll
ll
Klll
-1)mu(k,u(k),min
Vector notation:
[ ]{ })()1()|1(22
(kUkRkkY u
predy
kUΔΓ++−+Γ
Δmin
)
)()()|()|1( kdSkUSkkYMkkY predpred Δ+Δ+=+ s.t.du
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 147: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/147.jpg)
Least Squares Formulation
⎥⎥⎦
⎤
⎢⎢⎣
⎡ Δ−+⎥⎦
⎤⎢⎣
⎡Γ=Δ
⎥⎥⎦
⎤
⎢⎢⎣
⎡
ΓΓ
0)()|()1(
00
)( kdSkkYMkRI
kUS predy
u
uy - d
⎥⎦
⎤⎢⎣
⎡ +Γ≡
⎦⎣⎦⎣
0)|1( kkEp
y
⎤⎡ + )|1( kke⎦⎣ 0
O ifi d bl ( > )⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
+
+=+
)|(
)|1()|1(
kpke
kkekkEp M
Overspecified problem (m>p):
1
)|1()(1
kkESSSkU pyTy
TuuTuuyTy
Tu +ΓΓ⎟⎟
⎠
⎞⎜⎜⎝
⎛ΓΓ+ΓΓ=Δ
−
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 148: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/148.jpg)
Receding Horizon Implementation
[ ] )|1(00)(1
kkESSSIku pyTy
TuuTuuyTy
Tu +ΓΓ⎟⎟
⎠
⎞⎜⎜⎝
⎛ΓΓ+ΓΓ=Δ
−
L[ ] )|()( p⎟⎠
⎜⎝
Off-line computation (KMPC)p ( )
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 149: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/149.jpg)
Unconstrained MPC Algorithm
Do not vary manipulated inputs for n intervals (assume no disturbances - system is at equilibrium)
Initialize - measure output )0(y [ ]TTT yyY )0(ˆ,,)0(ˆ)0|0( K=
measure Δd(0), get new measurements
P di ti
[ ]))1(),1(ˆ( dy Δ
)1()1()1|1()1|( ΔΔ kdSkSkkYMkkY duSPrediction:
Correction:
)1()1()1|1()1|( −Δ+−Δ+−−=− kdSkuSkkYMkkY duS
( ))1|()(ˆ)1|()|( −−+−= kkykykkYkkY FIK
Compute reference trajectory error)()|()1()|1( kdSkkYMkRkkEp Δ−−+=+
d
Control computation )|1()( kkEKku pMPC +=Δ
PASI 2008 Lecture Notes © Francis J. Doyle III
150Obtain measurement 3 goto set ,1)),1(),1(ˆ( +=+Δ+ kkkdky
![Page 150: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/150.jpg)
Tuning MPC
• Horizons (m,p)
• Penalty weights
• Filters ( )ff
( )uy ΓΓ ,
• Filters
– unmeasured disturbancef
( )ynff ,,1 K
– setpoint filter:
kkRMkkR −−=− )1|1()1|(( )
{ }ynff
kkRkRkkRkkR′′′′=′′
−−′′+−=,,
)1|()()1|()|(
1 K diag KIKI
F
F
PASI 2008 Lecture Notes © Francis J. Doyle III
y
![Page 151: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/151.jpg)
Constrained MPC
• General Structure of QP
xgHxx TT − minx
gradient vectorHessian
cCx ≥ s.t.x
inequality constraintinequality constraint
equation matrix
q yequation vector
• Several robust and reliable QP solvers available
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 152: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/152.jpg)
MPC Constraints
⎥⎤
⎢⎡
⎥⎤
⎢⎡ 00I L
Input Magnitude Input Rate
⎥⎥⎥⎥⎥⎤
⎢⎢⎢⎢⎢⎡
−+−−
−−
≥Δ⎥⎥⎥⎥⎥⎥
⎢⎢⎢⎢⎢⎢
⎤⎡−
⎥⎥⎥⎥
⎦⎢⎢⎢⎢
⎣)1()(
)1()1(
)()1(
)(00
0
kukumkuku
kuku
kUI
IIII M
L
L
OK
MOOM
⎥⎥⎥⎥⎥⎤
⎢⎢⎢⎢⎢⎡
Δ−+Δ−
Δ−
≥Δ⎥⎦
⎤⎢⎣
⎡−)(
)1(
)(
)(kumku
ku
kUII
M
⎥⎥⎥⎥
⎦⎢⎢⎢⎢
⎣ −−−+
−−
⎥⎥⎥⎥⎥⎥
⎦⎢⎢⎢⎢⎢⎢
⎣⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−
−
)1()1(
)1()(
0
00
kumku
kuku
IIII
IM
L
OK
MOOM
L
⎥⎥⎥⎥
⎦⎢⎢⎢⎢
⎣ −+Δ−
Δ−⎥⎦
⎢⎣
)1(
)(
mku
kuIM
⎦⎣
⎤⎡ ⎤⎡
Output Magnitude
⎥⎥⎥⎥⎥⎤
⎢⎢⎢⎢⎢⎡
⎤⎡ +
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
+
+−Δ+
≥Δ⎥⎥⎦
⎤
⎢⎢⎣
⎡−)1()(
)1()()|(
)(ky
pky
kykdSkkYM
kUSSu
u
M d
PASI 2008 Lecture Notes © Francis J. Doyle III
⎥⎥⎥⎥
⎦⎢⎢⎢⎢
⎣⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
+
++Δ−−
⎥⎦⎢⎣
)(
)1()()|(
pky
kykdSkkYM
SM
d
![Page 153: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/153.jpg)
Constrained Formulation
)()|1()()( kUkkGkUHkU TuT
ΔU(k)+− ΔΔΔ min
)|1()( kkckUC uu +≥Δ s.t.
SSH uTuuyTyT
uu ΓΓ+ΓΓ=
)|1( kkESSG uyTyT
uu +ΓΓ )|1( kkESSG pyy +ΓΓ=
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 154: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/154.jpg)
Important Observations
• An unconstrained model predictive controller can be recast as a classical controller (PID, etc.) – First-order dynamic model = PIy– Second-order dynamic model = PID– More complex models yield more complex controllers
• MPC is a very general framework and represents the state of the art in a number of commercial sectors (refining, chemicals, aerospace, etc.)
• Strategy: understanding influences control design
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 155: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/155.jpg)
MPC Approach for Biomedical Control
• Many characteristics and requirements in common with industrial process control
• Successful drug delivery (clinical) studies – atracurium (Linkens and Mahfouf, 1995)( )– sodium nitroprusside (Kwok et al., 1997)– sodium nitroprusside & dopamine (Rao et al., 1999)– anesthesia (Gentilini et al., 2001)
• Present studies – computer patient models – Bergman Model (Bergman et al., 1981)g ( g )– Sorensen Model (Sorensen & Colton, 1985)– AIDA Model (Lehmann & Deutsch, 1992)
PASI 2008 Lecture Notes © Francis J. Doyle III
![Page 156: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/156.jpg)
Insulin Delivery – Algorithmic Details
• Solves on-line optimization problem• Controller objective function:
2
2
2
2)()|1()|1( kUkkRkkY uy ΔΓ++−+Γ
)(min
kUΔ
Controller tuning
Glucose Tracking
Insulin Penalty
• Controller tuning» move horizon, m, and prediction horizon, p» setpoint tracking (Γy), move suppression (Γu) weighting
• Constraints: 0 66 2516 5
mU min U k mU minU k mU min/ ( ) . /
( ) /≤ ≤
≤ΔPump
Limitations
PASI 2008 Lecture Notes © Francis J. Doyle III
16 560
U k mU minG k mg dl
( ) . /( ) /min
≤
≥
Δ
Safety
![Page 157: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/157.jpg)
100
150
(mg/
dL)
50
100la
sma
Glu
cose
Glucose [mg/dL]Hypoglycemic boundary
0 100 200 300 400 500 600 700 800 900 10000
Pl
25
yp g y ySetpoint
15
20
25
fusi
on (U
/h)
0 100 200 300 400 500 600 700 800 900 10000
5
10
Insu
lin In
Simple MPC with no meal detection. 15 units of insulin were used to cover the meal (lower figure in black), which led to a severe hypoglycemia. The setpoint denoted by the red dotted line, plasma glucose
0 100 200 300 400 500 600 700 800 900 1000Time (min)
PASI 2008 Lecture Notes © Francis J. Doyle III
), yp g y p y , p gwith blue line, hypoglycemic boundary (70 mg/dL) with green dashed line and controller moves with the black line in the lower plot
![Page 158: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/158.jpg)
150
200
(mg/
dL)
Glucose [mg/dL]Hypoglycemic boundarySetpoint
100
150as
ma
Glu
cose
Meal flag
0 100 200 300 400 500 600 700 800 900 100050
Pl
20
10
15
20
usio
n (U
/h)
0
5
10
Insu
lin In
fu
MPC with meal detection and variable reference. 11.5 units of insulin were used to cover the meal, which led to a mild hypoglycemia. The setpoint denoted by the red dotted line, plasma glucose with blue line, hypoglycemic
0 100 200 300 400 500 600 700 800 900 10000
Time (min)
PASI 2008 Lecture Notes © Francis J. Doyle III
boundary (70 mg/dL) with green dashed line, meal flag point with the green circle and controller moves with the black line in the lower plot.
![Page 159: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/159.jpg)
120
140
160
e (m
g/dL
)
Glucose [mg/dL]Hypoglycemic boundarySetpointMeal flag
80
100
120
Pla
sma
Glu
cose Meal flag
20
0 100 200 300 400 500 600 700 800 900 100060
P
10
15
20
fusi
on (U
/h)
0
5
Insu
lin In
f
0 100 200 300 400 500 600 700 800 900 1000
Time (min)
MPC with meal detection, variable reference and estimated glucose absorption profile and process noise of ± 3 mg/dL. 11 units of insulin were used to cover the meal. The setpoint denoted by the red dotted
PASI 2008 Lecture Notes © Francis J. Doyle III
line, plasma glucose with blue line, hypoglycemic boundary (70 mg/dL) with green dashed line, meal flag point with the green circle and controller moves with the black line in the lower plot.
![Page 160: Overview – Systems BiologySystems Biologycepac.cheme.cmu.edu/pasi2008/slides/doyle/library/... · • Will study: proteome is dynamic – changing w/ environment – But promoters](https://reader034.vdocuments.site/reader034/viewer/2022042219/5ec5156de03fd12955497756/html5/thumbnails/160.jpg)
140
160
(mg/
dL)
Glucose [mg/dL]Hypoglycemic boundarySetpoint
80
100
120as
ma
Glu
cose
p
0 100 200 300 400 500 600 700 800 900 100060
Pl
30
15
20
25
30
fusi
on (U
/h)
0 100 200 300 400 500 600 700 800 900 1000
5
10
15
Insu
lin In
f
MPC with announced meal variable reference and estimated glucose absorption profile 10 units of insulin were
0 100 200 300 400 500 600 700 800 900 1000Time (min)
PASI 2008 Lecture Notes © Francis J. Doyle III
MPC with announced meal, variable reference and estimated glucose absorption profile. 10 units of insulin were used to cover the meal. The setpoint denoted by the red dotted line, plasma glucose with blue line, hypoglycemic boundary (70 mg/dL) with green dashed line, meal flag point with the green circle and controller moves with the black line in the lower plot.