![Page 1: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/1.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Embedding Machine Learning inFormal Stochastic Models of Biological Processes
Jane Hillston
School of Informatics, University of Edinburgh
28th November 2017
quan col. . ...............................
![Page 2: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/2.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The PEPA project
The PEPA project started in Edinburgh in 1991.
It was motivated by problems encountered when carrying outperformance analysis of large computer and communicationsystems, based on numerical analysis of Markov chains.
Process algebras offered a compositional description techniquesupported by apparatus for formal reasoning.
Performance Evaluation Process Algebra (PEPA) sought toaddress these problems by the introduction of a suitableprocess algebra.
We have sought to investigate and exploit the interplaybetween the process algebra and the continuous time Markovchain (CTMC)
![Page 3: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/3.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The PEPA project
The PEPA project started in Edinburgh in 1991.
It was motivated by problems encountered when carrying outperformance analysis of large computer and communicationsystems, based on numerical analysis of Markov chains.
Process algebras offered a compositional description techniquesupported by apparatus for formal reasoning.
Performance Evaluation Process Algebra (PEPA) sought toaddress these problems by the introduction of a suitableprocess algebra.
We have sought to investigate and exploit the interplaybetween the process algebra and the continuous time Markovchain (CTMC)
![Page 4: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/4.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The PEPA project
The PEPA project started in Edinburgh in 1991.
It was motivated by problems encountered when carrying outperformance analysis of large computer and communicationsystems, based on numerical analysis of Markov chains.
Process algebras offered a compositional description techniquesupported by apparatus for formal reasoning.
Performance Evaluation Process Algebra (PEPA) sought toaddress these problems by the introduction of a suitableprocess algebra.
We have sought to investigate and exploit the interplaybetween the process algebra and the continuous time Markovchain (CTMC)
![Page 5: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/5.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The PEPA project
The PEPA project started in Edinburgh in 1991.
It was motivated by problems encountered when carrying outperformance analysis of large computer and communicationsystems, based on numerical analysis of Markov chains.
Process algebras offered a compositional description techniquesupported by apparatus for formal reasoning.
Performance Evaluation Process Algebra (PEPA) sought toaddress these problems by the introduction of a suitableprocess algebra.
We have sought to investigate and exploit the interplaybetween the process algebra and the continuous time Markovchain (CTMC)
![Page 6: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/6.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The PEPA project
The PEPA project started in Edinburgh in 1991.
It was motivated by problems encountered when carrying outperformance analysis of large computer and communicationsystems, based on numerical analysis of Markov chains.
Process algebras offered a compositional description techniquesupported by apparatus for formal reasoning.
Performance Evaluation Process Algebra (PEPA) sought toaddress these problems by the introduction of a suitableprocess algebra.
We have sought to investigate and exploit the interplaybetween the process algebra and the continuous time Markovchain (CTMC)
![Page 7: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/7.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 8: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/8.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 9: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/9.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Process Algebra
Models consist of agents which engage in actions.
α.P���*
HHHY
action typeor name
agent/component
The structured operational (interleaving) semantics of thelanguage is used to generate a labelled transition system.
Process algebra model Labelled transition system-SOS rules
![Page 10: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/10.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Process Algebra
Models consist of agents which engage in actions.
α.P���*
HHHY
action typeor name
agent/component
The structured operational (interleaving) semantics of thelanguage is used to generate a labelled transition system.
Process algebra model Labelled transition system-SOS rules
![Page 11: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/11.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic process algebras
Stochastic process algebra
Process algebras where models are decorated with quantitativeinformation used to generate a stochastic process are stochasticprocess algebras (SPA).
![Page 12: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/12.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic Process Algebra
Models are constructed from components which engage inactivities.
(α, r).P
���* 6 HH
HY
action typeor name
activity rate(parameter of an
exponential distribution)
component/derivative
The language is used to generate a Continuous Time MarkovChain (CTMC) for performance modelling.
SPAMODEL
LABELLEDMULTI-
TRANSITIONSYSTEM
CTMC Q- -SOS rules state transition
diagram
The CTMC can be analysed numerically (linear algebra) or bystochastic simulation.
![Page 13: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/13.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic Process Algebra
Models are constructed from components which engage inactivities.
(α, r).P
���* 6 HH
HY
action typeor name
activity rate(parameter of an
exponential distribution)
component/derivative
The language is used to generate a Continuous Time MarkovChain (CTMC) for performance modelling.
SPAMODEL
LABELLEDMULTI-
TRANSITIONSYSTEM
CTMC Q- -SOS rules state transition
diagram
The CTMC can be analysed numerically (linear algebra) or bystochastic simulation.
![Page 14: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/14.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic Process Algebra
Models are constructed from components which engage inactivities.
(α, r).P
���* 6 HH
HY
action typeor name
activity rate(parameter of an
exponential distribution)
component/derivative
The language is used to generate a Continuous Time MarkovChain (CTMC) for performance modelling.
SPAMODEL
LABELLEDMULTI-
TRANSITIONSYSTEM
CTMC Q- -SOS rules state transition
diagram
The CTMC can be analysed numerically (linear algebra) or bystochastic simulation.
![Page 15: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/15.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Integrated analysis
Qualitative verification can now be complemented by quantitativeverification.
Reachability analysisSpecification matchingModel checking
How long will it takefor the system to arrive
in a particular state?
e ee e e eiee e- - -
?����
���
-
���With what probabilitydoes system behaviourmatch its specification?
ee e e
e-
6
-
?
����
∼=?
e ee e e eee e- - -
?����
���
-
���Does a given property φhold within the system
with a given probability?
For a given starting statehow long is it until
a given property φ holds?φ �
����
���
PPPPPPPP
e ee e e eee e- - -
?����
���
-
���
![Page 16: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/16.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Integrated analysis
Qualitative verification can now be complemented by quantitativeverification.
Reachability analysis
Specification matchingModel checking
How long will it takefor the system to arrive
in a particular state?
e ee e e eiee e- - -
?����
���
-
���
With what probabilitydoes system behaviourmatch its specification?
ee e e
e-
6
-
?
����
∼=?
e ee e e eee e- - -
?����
���
-
���Does a given property φhold within the system
with a given probability?
For a given starting statehow long is it until
a given property φ holds?φ �
����
���
PPPPPPPP
e ee e e eee e- - -
?����
���
-
���
![Page 17: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/17.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Integrated analysis
Qualitative verification can now be complemented by quantitativeverification.
Reachability analysis
Specification matching
Model checking
How long will it takefor the system to arrive
in a particular state?
e ee e e eiee e- - -
?����
���
-
���
With what probabilitydoes system behaviourmatch its specification?
ee e e
e-
6
-
?
����
∼=?
e ee e e eee e- - -
?����
���
-
���
Does a given property φhold within the system
with a given probability?
For a given starting statehow long is it until
a given property φ holds?φ �
����
���
PPPPPPPP
e ee e e eee e- - -
?����
���
-
���
![Page 18: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/18.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Integrated analysis
Qualitative verification can now be complemented by quantitativeverification.
Reachability analysisSpecification matching
Model checking
How long will it takefor the system to arrive
in a particular state?
e ee e e eiee e- - -
?����
���
-
���With what probabilitydoes system behaviourmatch its specification?
ee e e
e-
6
-
?
����
∼=?
e ee e e eee e- - -
?����
���
-
���
Does a given property φhold within the system
with a given probability?
For a given starting statehow long is it until
a given property φ holds?
φ ���
�����
PPPPPPPP
e ee e e eee e- - -
?����
���
-
���
![Page 19: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/19.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Integrated analysis
Qualitative verification can now be complemented by quantitativeverification.
Reachability analysisSpecification matching
Model checking
How long will it takefor the system to arrive
in a particular state?
e ee e e eiee e- - -
?����
���
-
���With what probabilitydoes system behaviourmatch its specification?
ee e e
e-
6
-
?
����
∼=?
e ee e e eee e- - -
?����
���
-
���Does a given property φhold within the system
with a given probability?
For a given starting statehow long is it until
a given property φ holds?φ �
����
���
PPPPPPPP
e ee e e eee e- - -
?����
���
-
���
![Page 20: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/20.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Benefits of integration
Properties of the underlying mathematical structure may bededuced by the construction at the process algebra level.
Compositionality can be exploited both for model constructionand (in some cases) for model analysis.
Formal reasoning techniques such as equivalence relations andmodel checking can be used to manipulate or interrogatemodels.
For example the congruence Markovian bisimulation, allowsexact model reduction to be carried out compositionally.
Stochastic model checking based on the ContinuousStochastic Logic (CSL) allows automatic evaluation ofquantified properties of the behaviour of the system.
![Page 21: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/21.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Benefits of integration
Properties of the underlying mathematical structure may bededuced by the construction at the process algebra level.
Compositionality can be exploited both for model constructionand (in some cases) for model analysis.
Formal reasoning techniques such as equivalence relations andmodel checking can be used to manipulate or interrogatemodels.
For example the congruence Markovian bisimulation, allowsexact model reduction to be carried out compositionally.
Stochastic model checking based on the ContinuousStochastic Logic (CSL) allows automatic evaluation ofquantified properties of the behaviour of the system.
![Page 22: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/22.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Benefits of integration
Properties of the underlying mathematical structure may bededuced by the construction at the process algebra level.
Compositionality can be exploited both for model constructionand (in some cases) for model analysis.
Formal reasoning techniques such as equivalence relations andmodel checking can be used to manipulate or interrogatemodels.
For example the congruence Markovian bisimulation, allowsexact model reduction to be carried out compositionally.
Stochastic model checking based on the ContinuousStochastic Logic (CSL) allows automatic evaluation ofquantified properties of the behaviour of the system.
![Page 23: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/23.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Benefits of integration
Properties of the underlying mathematical structure may bededuced by the construction at the process algebra level.
Compositionality can be exploited both for model constructionand (in some cases) for model analysis.
Formal reasoning techniques such as equivalence relations andmodel checking can be used to manipulate or interrogatemodels.
For example the congruence Markovian bisimulation, allowsexact model reduction to be carried out compositionally.
Stochastic model checking based on the ContinuousStochastic Logic (CSL) allows automatic evaluation ofquantified properties of the behaviour of the system.
![Page 24: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/24.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Benefits of integration
Properties of the underlying mathematical structure may bededuced by the construction at the process algebra level.
Compositionality can be exploited both for model constructionand (in some cases) for model analysis.
Formal reasoning techniques such as equivalence relations andmodel checking can be used to manipulate or interrogatemodels.
For example the congruence Markovian bisimulation, allowsexact model reduction to be carried out compositionally.
Stochastic model checking based on the ContinuousStochastic Logic (CSL) allows automatic evaluation ofquantified properties of the behaviour of the system.
![Page 25: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/25.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 26: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/26.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic process algebras
Over the last two decades stochastic process algebras (mostly withMarkovian semantics) have been applied to a wide range ofapplication domains.
In some case there have been new languages developed to supportparticular features of the application domain. These have includedstochastic process algebras for modelling hybrid systems(e.g. HYPE, HyPA), spatial temporal systems (e.g. PALOMA,CARMA) and ecological processes (e.g. PALPS, MELA).
This is most noticeable in the arena of systems biology, which isoften focussed on biomolecular processing systems, for exampleBio-PEPA.
![Page 27: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/27.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic process algebras
Over the last two decades stochastic process algebras (mostly withMarkovian semantics) have been applied to a wide range ofapplication domains.
In some case there have been new languages developed to supportparticular features of the application domain. These have includedstochastic process algebras for modelling hybrid systems(e.g. HYPE, HyPA), spatial temporal systems (e.g. PALOMA,CARMA) and ecological processes (e.g. PALPS, MELA).
This is most noticeable in the arena of systems biology, which isoften focussed on biomolecular processing systems, for exampleBio-PEPA.
![Page 28: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/28.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Stochastic process algebras
Over the last two decades stochastic process algebras (mostly withMarkovian semantics) have been applied to a wide range ofapplication domains.
In some case there have been new languages developed to supportparticular features of the application domain. These have includedstochastic process algebras for modelling hybrid systems(e.g. HYPE, HyPA), spatial temporal systems (e.g. PALOMA,CARMA) and ecological processes (e.g. PALPS, MELA).
This is most noticeable in the arena of systems biology, which isoften focussed on biomolecular processing systems, for exampleBio-PEPA.
![Page 29: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/29.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Molecular processes as concurrent computations
ConcurrencyMolecularBiology
Metabolism SignalTransduction
Concurrentcomputational processes
Molecules Enzymes andmetabolites
Interactingproteins
Synchronous communication Molecularinteraction
Binding andcatalysis
Binding andcatalysis
Transition or mobilityBiochemicalmodification orrelocation
Metabolitesynthesis
Protein binding,modification orsequestration
A. Regev and E. Shapiro Cells as computation, Nature 419, 2002.
![Page 30: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/30.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The Bio-PEPA abstraction
Each “species” (biochemical component) i is described by aspecies component Ci
Each reaction j is associated with an action type αj and itsdynamics is described by a specific function fαj
The species components are then composed together to describethe behaviour of the system.
![Page 31: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/31.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The Bio-PEPA abstraction
Each “species” (biochemical component) i is described by aspecies component Ci
Each reaction j is associated with an action type αj and itsdynamics is described by a specific function fαj
The species components are then composed together to describethe behaviour of the system.
![Page 32: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/32.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The Bio-PEPA abstraction
Each “species” (biochemical component) i is described by aspecies component Ci
Each reaction j is associated with an action type αj and itsdynamics is described by a specific function fαj
The species components are then composed together to describethe behaviour of the system.
![Page 33: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/33.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bio-PEPA modelling
The state of the system at any time consists of the localstates of each of its sequential/species components.
The local states of components are quantitative rather thanfunctional, i.e. biological changes to species are represented asdistinct components.
A component varying its state corresponds to it varying itsamount.
This is captured by an integer parameter associated with thespecies and the effect of a reaction is to vary that parameterby a number corresponding to the stoichiometry of thisspecies in the reaction.
![Page 34: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/34.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bio-PEPA modelling
The state of the system at any time consists of the localstates of each of its sequential/species components.
The local states of components are quantitative rather thanfunctional, i.e. biological changes to species are represented asdistinct components.
A component varying its state corresponds to it varying itsamount.
This is captured by an integer parameter associated with thespecies and the effect of a reaction is to vary that parameterby a number corresponding to the stoichiometry of thisspecies in the reaction.
![Page 35: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/35.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bio-PEPA modelling
The state of the system at any time consists of the localstates of each of its sequential/species components.
The local states of components are quantitative rather thanfunctional, i.e. biological changes to species are represented asdistinct components.
A component varying its state corresponds to it varying itsamount.
This is captured by an integer parameter associated with thespecies and the effect of a reaction is to vary that parameterby a number corresponding to the stoichiometry of thisspecies in the reaction.
![Page 36: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/36.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bio-PEPA modelling
The state of the system at any time consists of the localstates of each of its sequential/species components.
The local states of components are quantitative rather thanfunctional, i.e. biological changes to species are represented asdistinct components.
A component varying its state corresponds to it varying itsamount.
This is captured by an integer parameter associated with thespecies and the effect of a reaction is to vary that parameterby a number corresponding to the stoichiometry of thisspecies in the reaction.
![Page 37: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/37.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 38: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/38.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 39: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/39.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 40: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/40.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 41: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/41.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 42: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/42.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 43: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/43.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 44: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/44.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 45: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/45.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The syntax
Sequential component (species component)
S ::= (α, κ) op S | S + S | C where op = ↓ | ↑ | ⊕ | | �
Model component
P ::= P BCLP | S(l)
Each action αj is associated with a rate fαj
![Page 46: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/46.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
The semantics
The semantics is defined by two transition relations:
First, a capability relation — is a transition possible?;
Second, a stochastic relation — gives rate of a transition,derived from the parameters of the model.
![Page 47: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/47.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Example
I RSIS S
Rspread
stop1stop2
k_s = 0.5;
k_r = 0.1;
kineticLawOf spread : k_s * I * S;
kineticLawOf stop1 : k_r * S * S;
kineticLawOf stop2 : k_r * S * R;
I = (spread,1) ↓ ;
S = (spread,1) ↑ + (stop1,1) ↓ + (stop2,1) ↓ ;
R = (stop1,1) ↑ + (stop2,1) ↑ ;
I[10] BC∗
S[5] BC∗
R[0]
![Page 48: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/48.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Example
I RSIS S
Rspread
stop1stop2
k_s = 0.5;
k_r = 0.1;
kineticLawOf spread : k_s * I * S;
kineticLawOf stop1 : k_r * S * S;
kineticLawOf stop2 : k_r * S * R;
I = (spread,1) ↓ ;
S = (spread,1) ↑ + (stop1,1) ↓ + (stop2,1) ↓ ;
R = (stop1,1) ↑ + (stop2,1) ↑ ;
I[10] BC∗
S[5] BC∗
R[0]
![Page 49: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/49.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Formal modelling in systems biology
Formal languages like Bio-PEPA provide a convenientinterface for describing complex systems, reflecting what isknow about the biological components and their behaviour.
High-level abstraction eases writing and manipulating models.
They are compiled into executable models which can be runto deepen understanding of the model.
Formal nature lends itself to automatic, rigorous methods foranalysis and verification.
Executing the model generates data that can be comparedwith biological data.
. . . but what if parts of the system are unknown?
Jasmin Fisher, Thomas A. Henzinger: Executable cell biology. Nature Biotechnology 2007
![Page 50: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/50.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Optimizing models
Usual process of parameterising a model is iterative and manual.
model
data
simulate/analyse
update
![Page 51: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/51.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Alternative perspective
?
?
Model creation is data-driven
![Page 52: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/52.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Machine Learning: Bayesian statistics
prior
posterior
data
inference
Represent belief and uncertainty as probability distributions(prior, posterior).
Treat parameters and unobserved variables similarly.
Bayes’ Theorem:
P(θ | D) =P(θ) · P(D | θ)
P(D)
posterior ∝ prior · likelihood
![Page 53: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/53.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Machine Learning: Bayesian statistics
prior
posterior
data
inference
Represent belief and uncertainty as probability distributions(prior, posterior).
Treat parameters and unobserved variables similarly.
Bayes’ Theorem:
P(θ | D) =P(θ) · P(D | θ)
P(D)
posterior ∝ prior · likelihood
![Page 54: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/54.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bayesian statistics
0 5 10 150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Prior distribution (no data)
![Page 55: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/55.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bayesian statistics
0 5 10 150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
After 20 data points
![Page 56: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/56.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bayesian statistics
0 5 10 150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
After 40 data points
![Page 57: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/57.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Bayesian statistics
0 5 10 150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
After 60 data points
![Page 58: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/58.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Modelling
Thus there are two approaches to model construction:
Machine Learning: extracting a model from the data generated bythe system, or refining a model based on systembehaviour using statistical techniques.
Mechanistic Modelling: starting from a description or hypothesis,construct a formal model that algorithmically mimicsthe behaviour of the system, validated against data.
![Page 59: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/59.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Comparing the techniques
Data-driven modelling:
+ rigorous handling of parameter uncertainty- limited or no treatment of stochasticity- in many cases bespoke solutions are required
which can limit the size of system which can behandled
Mechanistic modelling:
+ general execution ”engine” (deterministic orstochastic) can be reused for many models
+ models can be used speculatively to investigateroles of parameters, or alternative hypotheses
- parameters are assumed to be known and fixed,or costly approaches must be used to seekappropriate parameterisation
![Page 60: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/60.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Comparing the techniques
Data-driven modelling:
+ rigorous handling of parameter uncertainty- limited or no treatment of stochasticity- in many cases bespoke solutions are required
which can limit the size of system which can behandled
Mechanistic modelling:
+ general execution ”engine” (deterministic orstochastic) can be reused for many models
+ models can be used speculatively to investigateroles of parameters, or alternative hypotheses
- parameters are assumed to be known and fixed,or costly approaches must be used to seekappropriate parameterisation
![Page 61: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/61.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Developing a probabilistic programming approach
What if we could...
include information about uncertainty in the model?
automatically use observations to refine this uncertainty?
do all this in a formal context?
Starting from the existing process algebra (Bio-PEPA), we havedeveloped a new language ProPPA that addresses these issues.
A.Georgoulas, J.Hillston, D.Milios, G.Sanguinetti: Probabilistic Programming Process Algebra. QEST 2014.
![Page 62: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/62.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 63: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/63.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Probabilistic programming
A programming paradigm for describing incomplete knowledgescenarios, and resolving the uncertainty.
Programs probabilistic models in a high level language, likesoftware code.
Offers automated inference without the need to write bespokesolutions.
Platforms: IBAL, Church, Infer.NET, Fun, Anglican,WebPPL,....
Key actions: specify a distribution, specify observations, inferposterior distribution.
![Page 64: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/64.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Probabilistic programming workflow
Describe how the data is generated in syntax like aconventional programming language, but leaving somevariables uncertain.
Specify observations, which impose constraints on acceptableoutputs of the program.
Run program forwards: Generate data consistent withobservations.
Run program backwards: Find values for the uncertainvariables which make the output match the observations.
![Page 65: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/65.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Probabilistic programming workflow
Describe how the data is generated in syntax like aconventional programming language, but leaving somevariables uncertain.
Specify observations, which impose constraints on acceptableoutputs of the program.
Run program forwards: Generate data consistent withobservations.
Run program backwards: Find values for the uncertainvariables which make the output match the observations.
![Page 66: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/66.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Probabilistic programming workflow
Describe how the data is generated in syntax like aconventional programming language, but leaving somevariables uncertain.
Specify observations, which impose constraints on acceptableoutputs of the program.
Run program forwards: Generate data consistent withobservations.
Run program backwards: Find values for the uncertainvariables which make the output match the observations.
![Page 67: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/67.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Probabilistic programming workflow
Describe how the data is generated in syntax like aconventional programming language, but leaving somevariables uncertain.
Specify observations, which impose constraints on acceptableoutputs of the program.
Run program forwards: Generate data consistent withobservations.
Run program backwards: Find values for the uncertainvariables which make the output match the observations.
![Page 68: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/68.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
A Probabilistic Programming Process Algebra: ProPPA
The objective of ProPPA is to retain the features of the stochasticprocess algebra:
simple model description in terms of components
rigorous semantics giving an executable version of the model...
... whilst also incorporating features of a probabilistic programminglanguage:
recording uncertainty in the parameters
ability to incorporate observations into models
access to inference to update uncertainty based onobservations
![Page 69: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/69.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
A Probabilistic Programming Process Algebra: ProPPA
The objective of ProPPA is to retain the features of the stochasticprocess algebra:
simple model description in terms of components
rigorous semantics giving an executable version of the model...
... whilst also incorporating features of a probabilistic programminglanguage:
recording uncertainty in the parameters
ability to incorporate observations into models
access to inference to update uncertainty based onobservations
![Page 70: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/70.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Example Revisited
I RSIS S
Rspread
stop1stop2
k_s = 0.5;
k_r = 0.1;
kineticLawOf spread : k_s * I * S;
kineticLawOf stop1 : k_r * S * S;
kineticLawOf stop2 : k_r * S * R;
I = (spread,1) ↓ ;
S = (spread,1) ↑ + (stop1,1) ↓ + (stop2,1) ↓ ;
R = (stop1,1) ↑ + (stop2,1) ↑ ;
I[10] BC∗
S[5] BC∗
R[0]
![Page 71: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/71.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Additions
Declaring uncertain parameters:
k s = Uniform(0,1);
k t = Uniform(0,1);
Providing observations:
observe(’trace’)
Specifying inference approach:
infer(’ABC’)
![Page 72: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/72.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Additions
I RSIS S
Rspread
stop1stop2
k_s = Uniform(0,1);
k_r = Uniform(0,1);
kineticLawOf spread : k_s * I * S;
kineticLawOf stop1 : k_r * S * S;
kineticLawOf stop2 : k_r * S * R;
I = (spread,1) ↓ ;
S = (spread,1) ↑ + (stop1,1) ↓ + (stop2,1) ↓ ;
R = (stop1,1) ↑ + (stop2,1) ↑ ;
I[10] BC∗
S[5] BC∗
R[0]
observe(’trace’)
infer(’ABC’) //Approximate Bayesian Computation
![Page 73: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/73.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Semantics
A Bio-PEPA model can be interpreted as a CTMC; however,CTMCs cannot capture uncertainty in the rates (everytransition must have a concrete rate).
ProPPA models include uncertainty in the parameters, whichtranslates into uncertainty in the transition rates.
A ProPPA model should be mapped to something like adistribution over CTMCs.
![Page 74: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/74.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
parameter
model
k = 2
CTMC
![Page 75: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/75.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
parameter
model
k ∈ [0,5]
set of CTMCs
![Page 76: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/76.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
parameter
model
k ∼ p
distributionover CTMCs
μ
![Page 77: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/77.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Constraint Markov Chains
Constraint Markov Chains (CMCs) are a generalization of DTMCs,in which the transition probabilities are not concrete, but can takeany value satisfying some constraints.
Constraint Markov Chain
A CMC is a tuple 〈S , o,A,V , φ〉, where:
S is the set of states, of cardinality k.
o ∈ S is the initial state.
A is a set of atomic propositions.
V : S → 22A
gives a set of acceptable labellings for each state.
φ : S × [0, 1]k → {0, 1} is the constraint function.
Caillaud et al., Constraint Markov Chains, Theoretical Computer Science, 2011
![Page 78: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/78.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Constraint Markov Chains
In a CMC, arbitrary constraints are permitted, expressed throughthe function φ: φ(s, ~p) = 1 iff ~p is an acceptable vector oftransition probabilities from state s.
However,
CMCs are defined only for the discrete-time case, and
this does not say anything about how likely a value is to bechosen, only about whether it is acceptable.
To address these shortcomings, we define ProbabilisticConstraint Markov Chains.
![Page 79: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/79.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
![Page 80: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/80.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Probabilistic CMCs
A Probabilistic Constraint Markov Chain is a tuple 〈S , o,A,V , φ〉,where:
S is the set of states, of cardinality k.
o ∈ S is the initial state.
A is a set of atomic propositions.
V : S → 22A
gives a set of acceptable labellings for each state.
φ : S × [0,∞)k → [0,∞) is the constraint function.
This is applicable to continuous-time systems.
φ(s, ·) is now a probability density function on the transitionrates from state s.
![Page 81: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/81.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Semantics of ProPPA
The semantics definition follows that of Bio-PEPA, which isdefined using two transition relations:
Capability relation — is a transition possible?
Stochastic relation — gives distribution of the rate of atransition
The distribution over the parameter values induces a distributionover transition rates.
Rules are expressed as state-to-function transition systems(FuTS1).
This gives rise the underlying PCMC.
1De Nicola et al., A Uniform Definition of Stochastic Process Calculi, ACM Computing Surveys, 2013
![Page 82: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/82.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Semantics of ProPPA
The semantics definition follows that of Bio-PEPA, which isdefined using two transition relations:
Capability relation — is a transition possible?
Stochastic relation — gives distribution of the rate of atransition
The distribution over the parameter values induces a distributionover transition rates.
Rules are expressed as state-to-function transition systems(FuTS1).
This gives rise the underlying PCMC.
1De Nicola et al., A Uniform Definition of Stochastic Process Calculi, ACM Computing Surveys, 2013
![Page 83: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/83.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Simulating Probabilistic Constraint Markov Chains
Probabilistic Constraint Markov Chains are open to two alternativedynamic interpretations:
1 Uncertain Markov Chains: For each trajectory, for eachuncertain transition rate, sample once at the start of the runand use that value throughout;
2 Imprecise Markov Chains: During each trajectory, each time atransition with an uncertain rate is encountered, sample avalue but then discard it and re-sample whenever thistransition is visited again.
Our current work is focused on the Uncertain Markov Chain case.
![Page 84: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/84.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Simulating Probabilistic Constraint Markov Chains
Probabilistic Constraint Markov Chains are open to two alternativedynamic interpretations:
1 Uncertain Markov Chains: For each trajectory, for eachuncertain transition rate, sample once at the start of the runand use that value throughout;
2 Imprecise Markov Chains: During each trajectory, each time atransition with an uncertain rate is encountered, sample avalue but then discard it and re-sample whenever thistransition is visited again.
Our current work is focused on the Uncertain Markov Chain case.
![Page 85: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/85.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 86: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/86.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inference
parameter
model
k ∼ p
distributionover CTMCs
μ
![Page 87: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/87.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inference
parameter
model
k ∼ p
distributionover CTMCs
μobservations
inference
posteriordistribution
μ*
![Page 88: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/88.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inference
P(θ | D) ∝ P(θ)P(D | θ)
Exact inference is impossible, as we cannot calculate thelikelihood.
We must use approximate algorithms or approximations of thesystem.
The ProPPA semantics does not define a single inferencealgorithm, allowing for a modular approach.
Different algorithms can act on different input (time-series vsproperties), return different results or in different forms.
![Page 89: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/89.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inferring likelihood in uncertain CTMCs
Transient state probabilities can be expressed as:
dpi (t)
dt=
∑j 6=i
pj(t) · qji − pi (t)∑j 6=i
qij
The probability of a single observation (y , t) can then be expressedas
p(y , t) =∑i∈S
pi (t)π(y | i)
where π(y | i) is the probability of observing y when in state i .
The likelihood can then be expressed as
P(D | θ) =N∏j=1
∑i∈S
p(i |θ)(tj)π(yj | i)
![Page 90: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/90.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inferring likelihood in uncertain CTMCs
Transient state probabilities can be expressed as:
dpi (t)
dt=
∑j 6=i
pj(t) · qji − pi (t)∑j 6=i
qij
The probability of a single observation (y , t) can then be expressedas
p(y , t) =∑i∈S
pi (t)π(y | i)
where π(y | i) is the probability of observing y when in state i .
The likelihood can then be expressed as
P(D | θ) =N∏j=1
∑i∈S
p(i |θ)(tj)π(yj | i)
![Page 91: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/91.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inferring likelihood in uncertain CTMCs
Transient state probabilities can be expressed as:
dpi (t)
dt=
∑j 6=i
pj(t) · qji − pi (t)∑j 6=i
qij
The probability of a single observation (y , t) can then be expressedas
p(y , t) =∑i∈S
pi (t)π(y | i)
where π(y | i) is the probability of observing y when in state i .
The likelihood can then be expressed as
P(D | θ) =N∏j=1
∑i∈S
p(i |θ)(tj)π(yj | i)
![Page 92: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/92.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Calculating the transient probabilities
For finite state-spaces, the transient probabilities can, in principle,be computed as
p(t) = p(0)eQt .
Likelihood is hard to compute:
Computing eQt is expensive if the state space is large
Impossible directly in infinite state-spaces
![Page 93: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/93.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Basic Inference
Approximate Bayesian Computation is a simplesimulation-based solution:
Approximates posterior distribution over parameters as a set ofsamplesLikelihood of parameters is approximated with a notion ofdistance.
![Page 94: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/94.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Basic Inference
Approximate Bayesian Computation is a simplesimulation-based solution:
Approximates posterior distribution over parameters as a set ofsamplesLikelihood of parameters is approximated with a notion ofdistance.
x
xx x
t
X
![Page 95: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/95.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Basic Inference
Approximate Bayesian Computation is a simplesimulation-based solution:
Approximates posterior distribution over parameters as a set ofsamplesLikelihood of parameters is approximated with a notion ofdistance.
x
xx x
t
X
![Page 96: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/96.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Basic Inference
Approximate Bayesian Computation is a simplesimulation-based solution:
Approximates posterior distribution over parameters as a set ofsamplesLikelihood of parameters is approximated with a notion ofdistance.
x
xx x
t
X
![Page 97: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/97.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Basic Inference
Approximate Bayesian Computation is a simplesimulation-based solution:
Approximates posterior distribution over parameters as a set ofsamplesLikelihood of parameters is approximated with a notion ofdistance.
x
xx x
t
XΣ(xi-yi)
2 > εrejected
![Page 98: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/98.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Basic Inference
Approximate Bayesian Computation is a simplesimulation-based solution:
Approximates posterior distribution over parameters as a set ofsamplesLikelihood of parameters is approximated with a notion ofdistance.
x
xx x
t
XΣ(xi-yi)
2 < εaccepted
![Page 99: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/99.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Approximate Bayesian Computation
ABC algorithm
1 Sample a parameter set from the prior distribution.
2 Simulate the system using these parameters.
3 Compare the simulation trace obtained with the observations.
4 If distance < ε, accept, otherwise reject.
This results in an approximation to the posterior distribution.As ε→ 0, set of samples converges to true posterior.We use a more elaborate version based on Markov Chain MonteCarlo sampling.
![Page 100: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/100.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Inference for infinite state spaces
Various methods become inefficient or inapplicable as thestate-space grows.
How to deal with unbounded systems?
Multiple simulation runs
Large population approximations (diffusion, Linear Noise,. . . )
Systematic truncation
Random truncations
![Page 101: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/101.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Expanding the likelihood
The likelihood can be written as an infinite series:
p(x ′, t ′ | x , t) =∞∑
N=0
p(N)(x ′, t ′ | x , t)
=∞∑
N=0
[f (N)(x ′, t ′ | x , t)− f (N−1)(x ′, t ′ | x , t)
]where
x∗ = max{x , x ′}p(N)(x ′, t ′ | x , t) is the probability of going from state x attime t to state x ′ at time t ′ through a path with maximumstate x∗ + N
f (N) is the same, except the maximum state cannot exceedx∗ + N (but does not have to reach it)
Any finite number of terms can be computed — Can the infinitesum be computed or estimated?
![Page 102: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/102.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Expanding the likelihood
The likelihood can be written as an infinite series:
p(x ′, t ′ | x , t) =∞∑
N=0
p(N)(x ′, t ′ | x , t)
=∞∑
N=0
[f (N)(x ′, t ′ | x , t)− f (N−1)(x ′, t ′ | x , t)
]where
x∗ = max{x , x ′}p(N)(x ′, t ′ | x , t) is the probability of going from state x attime t to state x ′ at time t ′ through a path with maximumstate x∗ + N
f (N) is the same, except the maximum state cannot exceedx∗ + N (but does not have to reach it)
Any finite number of terms can be computed — Can the infinitesum be computed or estimated?
![Page 103: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/103.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
(0,0) (0,1) (0,3) (0,4)
(1,0) (1,3) (1,4)
(2,0) (2,1) (2,2) (2,3) (2,4)
(3,0) (3,1) (3,2) (3,3) (3,4)
(1,1)
(0,2)
(1,2)x
x'
![Page 104: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/104.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
(0,0) (0,1) (0,3) (0,4)
(1,0) (1,3) (1,4)
(2,0) (2,1) (2,2) (2,3) (2,4)
(3,0) (3,1) (3,2) (3,3) (3,4)
(1,1)
(0,2)
(1,2)
S0
x
x'
x*
![Page 105: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/105.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
(0,0) (0,1) (0,3) (0,4)
(1,0) (1,3) (1,4)
(2,0) (2,1) (2,2) (2,3) (2,4)
(3,0) (3,1) (3,2) (3,3) (3,4)
(1,1)
(0,2)
(1,2)
S0S1
x
x'
x*
![Page 106: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/106.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Russian Roulette Truncation
We want to estimate the value of
f =∞∑n=0
fn
where the fn’s are computable.
Truncate the sum randomly: stop at term k with probabilityqk .
Form f̂ as a partial sum of the fn, n = 1, . . . , k , rescaledappropriately.
f̂ =k∑
n=0
fnk−1∏j=0
(1− qj)
![Page 107: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/107.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Russian Roulette Truncation
We want to estimate the value of
f =∞∑n=0
fn
where the fn’s are computable.
Choose a single term fk with probability pk ; estimate f̂ = fkpk
f̂ is unbiased... but its variance can be high.
Truncate the sum randomly: stop at term k with probabilityqk .
Form f̂ as a partial sum of the fn, n = 1, . . . , k , rescaledappropriately.
f̂ =k∑
n=0
fnk−1∏j=0
(1− qj)
![Page 108: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/108.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Russian Roulette Truncation
We want to estimate the value of
f =∞∑n=0
fn
where the fn’s are computable.
Truncate the sum randomly: stop at term k with probabilityqk .
Form f̂ as a partial sum of the fn, n = 1, . . . , k , rescaledappropriately.
f̂ =k∑
n=0
fnk−1∏j=0
(1− qj)
![Page 109: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/109.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Russian Roulette
f̂ ← f0i ← 1p ← 1loop
Choose to stop with probability qiif stopping then
return f̂else
p ← p · (1− qi )f̂ ← f̂ + fi
pi ← i + 1
end ifend loop
![Page 110: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/110.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Russian Roulette
f̂ =k∑
n=0
fn∏k−1j=0 (1− qj)
In our case, f is a probability that we wish to approximate.
Using f̂ instead of f leads to an error; however f̂ is unbiased:E [f̂ ] = f .
f̂ is also guaranteed to be positive.
Pseudo-marginal algorithms can use this and still drawsamples from the correct distribution.
We have developed both Metropolis-Hastings and Gibbs-likesampling algorithms based on this approach2.
2Unbiased Bayesian Inference for Population Markov Jump Processes via Random Truncations. A.Georgoulas,
J.Hillston and D.Sanguinetti, in Stats & Comp. 2017
![Page 111: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/111.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
SSA samples
![Page 112: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/112.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Posterior samples
![Page 113: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/113.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 114: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/114.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Example model
I RSIS S
R
k_s = Uniform(0,1);
k_r = Uniform(0,1);
kineticLawOf spread : k_s * I * S;
kineticLawOf stop1 : k_r * S * S;
kineticLawOf stop2 : k_r * S * R;
I = (spread,1) ↓ ;
S = (spread,1) ↑ + (stop1,1) ↓ + (stop2,1) ↓ ;
R = (stop1,1) ↑ + (stop2,1) ↑ ;
I[10] BC∗
S[5] BC∗
R[0]
observe(’trace’)
infer(’ABC’) //Approximate Bayesian Computation
![Page 115: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/115.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Results
Tested on the rumour-spreading example, giving the twoparameters uniform priors.
Approximate Bayesian Computation
Returns posterior as a set of points (samples)
Observations: time-series (single simulation)
![Page 116: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/116.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Results: ABC
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
ks
k r
![Page 117: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/117.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Results: ABC
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
ks
k r
![Page 118: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/118.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Results: ABC
0 0.2 0.4 0.6 0.8 10
2000
4000
6000
8000
10000
12000
kr
Num
ber
of s
ampl
es
0 0.2 0.4 0.6 0.8 10
1000
2000
3000
4000
5000
6000
7000
ks
Num
ber
of s
ampl
es
![Page 119: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/119.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Genetic Toggle Switch
Two mutually-repressing genes: promoters (unobserved) andtheir protein products
Bistable behaviour: switching induced by environmentalchanges
Synthesised in E. coli3
Stochastic variant4 where switching is induced by noise
3Gardner, Cantor & Collins, Construction of a genetic toggle switch in Escherichia coli, Nature, 2000
4Tian & Burrage, Stochastic models for regulatory networks of the genetic toggle switch, PNAS, 2006
![Page 120: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/120.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Genetic Toggle Switch
P1
∅
∅P2
∅
∅
G1,on
G1,off G2,on
G2,off
participates accelerates
![Page 121: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/121.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Toggle switch model: species
G1 = activ1 ↑ + deact1 ↓ + expr1 ⊕;G2 = activ2 ↑ + deact2 ↓ + expr2 ⊕;
P1 = expr1 ↑ + degr1 ↓ + deact2 ⊕ ;
P2 = expr2 ↑ + degr2 ↓ + deact1 ⊕
G1[1] <*> G2[0] <*> P1[20] <*> P2[0]
observe(toggle_obs);
infer(rouletteGibbs);
![Page 122: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/122.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
θ 1 = Gamma(3,5); //etc...
kineticLawOf expr1 : θ 1 * G1;
kineticLawOf expr2 : θ 2 * G2;
kineticLawOf degr1 : θ 3 * P1;
kineticLawOf degr2 : θ 4 * P2;
kineticLawOf activ1 : θ 5 * (1 - G1);
kineticLawOf activ2 : θ 6 * (1 - G2);
kineticLawOf deact1 : θ 7 * exp(r ∗ P2) * G1;
kineticLawOf deact2 : θ 8 * exp(r ∗ P1) * G2;
G1 = activ1 ↑ + deact1 ↓ + expr1 ⊕;G2 = activ2 ↑ + deact2 ↓ + expr2 ⊕;P1 = expr1 ↑ + degr1 ↓ + deact2 ⊕ ;
P2 = expr2 ↑ + degr2 ↓ + deact1 ⊕
G1[1] <*> G2[0] <*> P1[20] <*> P2[0]
observe(toggle_obs);
infer(rouletteGibbs);
![Page 123: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/123.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Experiment
Simulated observations
Gamma priors on all parameters (required by algorithm)
Goal: learn posterior of 8 parameters
5000 samples taken using the Gibbs-like random truncationalgorithm
![Page 124: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/124.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Genes
0 20 40 60 80 100
![Page 125: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/125.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Proteins
0 20 40 60 80 1000
5
10
15
20
25
30
![Page 126: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/126.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Observations used
0 20 40 60 80 1000
5
10
15
20
25
30
![Page 127: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/127.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Results
![Page 128: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/128.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Outline
1 Stochastic Process Algebras
2 Modelling Biological ProcessesBio-PEPAApproaches to system biology modelling
3 ProPPA
4 Inference
5 Results
6 Conclusions
![Page 129: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/129.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Workflow
model
inference algorithm
low-level description
inference results
(samples)
statistics
plotting
prediction
...
infercompile
![Page 130: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/130.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Summary
ProPPA is a process algebra that incorporates uncertainty andobservations directly in the model, influenced by probabilisticprogramming.
Syntax remains similar to Bio-PEPA.
Semantics defined in terms of an extension of ConstraintMarkov Chains.
Observations can be either time-series or logical properties.
Parameter inference based on random truncations (RussianRoulette) offers new possibilities for inference.
![Page 131: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/131.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Challenges and Future Directions
The value of observations
Can we reason about the “distance” between µ and µ∗?
![Page 132: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/132.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Challenges and Future Directions
Heterogeneous populations
What if we are seeking the “optimal mix” rather than the bestindividual representative?
![Page 133: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/133.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Thanks
Bio-PEPA
FedericaCiocchetta
ProPPA
AnastasisGeorgoulas
GuidoSanguinetti
Funding from EPSRC, ERC, Microsoft Research and CECFET-Proactive Programme FoCAS.
![Page 134: Embedding Machine Learning in Formal Stochastic Models of ...homepages.inf.ed.ac.uk/jeh/TALKS/OxWoCS2017.pdf · Stochastic Process Algebras Modelling Biological ProcessesProPPAInferenceResultsConclusions](https://reader031.vdocuments.site/reader031/viewer/2022011902/5f0b825d7e708231d430deb0/html5/thumbnails/134.jpg)
Stochastic Process Algebras Modelling Biological Processes ProPPA Inference Results Conclusions
Thanks
Bio-PEPA
FedericaCiocchetta
ProPPA
AnastasisGeorgoulas
GuidoSanguinetti
Funding from EPSRC, ERC, Microsoft Research and CECFET-Proactive Programme FoCAS.