Michael P.H. Stumpf | Imperial College London (original) (raw)

Papers by Michael P.H. Stumpf

Research paper thumbnail of Information Geometry and Sequential Monte Carlo

arXiv preprint arXiv:1212.0764, Dec 4, 2012

Abstract: This paper explores the application of methods from information geometry to the sequent... more Abstract: This paper explores the application of methods from information geometry to the sequential Monte Carlo (SMC) sampler. In particular the Riemannian manifold Metropolis-adjusted Langevin algorithm (mMALA) is adapted for the transition kernels in SMC. Similar to its function in Markov chain Monte Carlo methods, the mMALA is a fully adaptable kernel which allows for efficient sampling of high-dimensional and highly correlated parameter spaces. We set up the theoretical framework for its use in SMC with a focus on the ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Goldstein-Kac telegraph processes with random speeds: Path probabilities, likelihoods, and reported Lévy flights

Physical review. E, Statistical, nonlinear, and soft matter physics, 2015

The Goldstein-Kac telegraph process describes the one-dimensional motion of particles with consta... more The Goldstein-Kac telegraph process describes the one-dimensional motion of particles with constant speed undergoing random changes in direction. Despite its resemblance to numerous real-world phenomena, the singular nature of the resultant spatial distribution of each particle precludes the possibility of any a posteriori empirical validation of this random-walk model from data. Here we show that by simply allowing for random speeds, the ballistic terms are regularized and that the diffusion component can be well-approximated via the unscented transform. The result is a computationally efficient yet robust evaluation of the full particle path probabilities and, hence, the parameter likelihoods of this generalized telegraph process. We demonstrate how a population diffusing under such a model can lead to non-Gaussian asymptotic spatial distributions, thereby mimicking the behavior of an ensemble of Lévy walkers.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Handbook of Statistical Systems Biology

Systems biology is data-rich. Technological advances over the past 50 years have allowed us to pr... more Systems biology is data-rich. Technological advances over the past 50 years have allowed us to probe, map and interfere with biological organisms in a number of ways. Most notable are perhaps the sequencing efforts that are continuing to catalogue the genetic diversity of life, including our own species. A whole host of other techniques from biochemistry, molecular, cell and structural biology have been used to study the function of the protein products and other biomolecules that are encoded by these sequences. So we live in a ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Statistical analysis of network data and evolution on GPUs: High-performance statistical computing

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Network Analysis Primer

Statistical and Evolutionary Analysis of …, Jan 1, 2010

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Inference of Signalling Pathway Models

Handbook of Statistical Systems …, Jan 1, 2011

Skip to Main Content. ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Considerate approaches to constructing summary statistics for ABC model selection

Statistics and Computing, Jan 1, 2012

Abstract For nearly any challenging scientific problem evaluation of the likelihood is problemati... more Abstract For nearly any challenging scientific problem evaluation of the likelihood is problematic if not impossible. Approximate Bayesian computation (ABC) allows us to employ the whole Bayesian formalism to problems where we can use simulations from a model, but cannot evaluate the likelihood directly. When summary statistics of real and simulated data are compared—rather than the data directly—information is lost, unless the summary statistics are sufficient. Sufficient statistics are, however, not common but without them ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Variable Drug Response: Genetic Evaluation

eLS, Jan 1, 2006

Pharmacogenetics emerged in the late 1950s and early 1960s when variation in the metabolism of th... more Pharmacogenetics emerged in the late 1950s and early 1960s when variation in the metabolism of the antimalarial agents primaquine and isoniazid and the antihypertensive debrisoquine was shown to follow a simple Mendelian inheritance pattern. Following such discoveries, Motulsky set the conceptual foundation for pharmacogenetics by pointing out that common nonclinical or subclinical genetic variation may underlie genetic variation contributing to pharmacokinetic differences among individuals. The classical debrisoquine ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Biosimulation in Drug Development

Human Genomics, Jan 1, 2008

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Statistical and Evolutionary Analysis of Biological Networks

Page 1. Statistical and Evolutionary Analysis of Biological Networks imperial CoJJege Press Page ... more Page 1. Statistical and Evolutionary Analysis of Biological Networks imperial CoJJege Press Page 2. Statistical and Evolutionary Analysis of Biological Networks Page 3. This page intentionally left blank This page intentionally left blank Page 4. Statistical and Evolutionary Analysis of Biological Networks Imperial College Press ICP editors Michael PH Stumpf Imperial College London, UK Carsten Wiuf Aarhus University, Denmark Page 5. Published ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Statistical Null Models for Biological Network Analysis

Statistical and Evolutionary …, Jan 1, 2010

Statistical ensembles of random graphs serve as null models in the statistical analysis of real c... more Statistical ensembles of random graphs serve as null models in the statistical analysis of real complex networks. They encapsulate what are believed to be the generic properties of networks and describe the expected behaviour against which observed network data can be compared. Here we review the basic statistical physics underlying statistical ensembles of networks and show how we can exploit their properties. We also show how the simple statistical ensembles that have been used to describe networks can be improved by ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Language's place in nature

Trends in Ecology & Evolution, Jan 1, 2001

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Binomial subsampling

Proceedings of the Royal …, Jan 1, 2006

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Critical truths about power laws

Science, Jan 1, 2012

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Bayesian design strategies for synthetic biology

Interface Focus, Jan 1, 2011

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Inference of Temporally Varying Bayesian Networks

Bioinformatics (Oxford, England), Jan 1, 2012

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Parameter-free model discrimination criterion based on steady-state coplanarity

Proceedings of the National Academy of Sciences of the United States of America, Jan 1, 2012

We introduce a procedure for deciding when a mass-action model is incompatible with observed stea... more We introduce a procedure for deciding when a mass-action model is incompatible with observed steady-state data that does not require any parameter estimation. Thus, we avoid the difficulties of nonlinear optimization typically associated with methods based on parameter fitting. Instead, we borrow ideas from algebraic geometry to construct a transformation of the model variables such that any set of steady states of the model under that transformation lies on a common plane, irrespective of the values of the model parameters. Model rejection can then be performed by assessing the degree to which the transformed data deviate from coplanarity. We demonstrate our method by applying it to models of multisite phosphorylation and cell death signaling. Our framework offers a parameter-free perspective on the statistical model selection problem, which can complement conventional statistical methods in certain classes of problems where inference has to be based on steady-state data and the model structures allow for suitable algebraic relationships among the steady-state solutions.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Evolutionary characteristics of bacterial two-component systems

Advances in experimental medicine and biology, Jan 1, 2012

The evolution of biological systems is influenced by a number of factors and forces that have act... more The evolution of biological systems is influenced by a number of factors and forces that have acted in different combinations at different times to give rise to extant organisms. Here we illustrate some of the issues surrounding the data-driven evolutionary analysis of biological systems in the context of bacterial two-component systems (TCSs). TCSs are critical for bacteria to interact with their extracellular environment. A typical TCS consists of a histidine kinase on the membrane and a response regulator in the cytoplasm. Here we comprehensively characterise the extent to which these appear together across some 950 bacterial species and test for statistically significant patterns of correlated gain and loss. Our analysis provides evidence for correlated evolution but also a high level of evolutionary flexibility: at the sequence level, histidine kinases but especially response regulators belonging to different TCSs in a species show high levels of similarity, which may facilitate crosstalk as well as the recruitment of components into new compound signalling systems. We furthermore find that bacterial lifestyle has an overriding influence on the presence and absence of TCS; while in most TCSs either both or none of the two components are present, several TCSs tend to lose preferentially either the histidine kinase or response regulator component, which further supports the notion of reuse and reshuffling of these components in different TCS arrangements. We conclude by placing these findings in a wider context and discuss the implications for evolutionary systems biology more generally.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Elucidating the in vivo phosphorylation dynamics of the ERK MAP kinase using quantitative proteomics data and Bayesian model selection

Molecular bioSystems, Jan 1, 2012

Ever since reversible protein phosphorylation was discovered, it has been clear that it plays a k... more Ever since reversible protein phosphorylation was discovered, it has been clear that it plays a key role in the regulation of cellular processes. Proteins often undergo double phosphorylation, which can occur through two possible mechanisms: distributive or processive. Which phosphorylation mechanism is chosen for a particular cellular regulation bears biological significance, and it is therefore in our interest to understand these mechanisms. In this paper we study dynamics of the MEK/ERK phosphorylation. We employ a model selection algorithm based on approximate Bayesian computation to elucidate phosphorylation dynamics from quantitative time course data obtained from PC12 cells in vivo. The algorithm infers the posterior distribution over four proposed models for phosphorylation and dephosphorylation dynamics, and this distribution indicates the amount of support given to each model. We evaluate the robustness of our inferential framework by systematically exploring different ways of parameterizing the models and for different prior specifications. The models with the highest inferred posterior probability are the two models employing distributive dephosphorylation, whereas we are unable to choose decisively between the processive and distributive phosphorylation mechanisms.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of ABC-SysBio--approximate Bayesian computation in Python with GPU support

Bioinformatics (Oxford, England), Jan 1, 2010

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Information Geometry and Sequential Monte Carlo

arXiv preprint arXiv:1212.0764, Dec 4, 2012

Abstract: This paper explores the application of methods from information geometry to the sequent... more Abstract: This paper explores the application of methods from information geometry to the sequential Monte Carlo (SMC) sampler. In particular the Riemannian manifold Metropolis-adjusted Langevin algorithm (mMALA) is adapted for the transition kernels in SMC. Similar to its function in Markov chain Monte Carlo methods, the mMALA is a fully adaptable kernel which allows for efficient sampling of high-dimensional and highly correlated parameter spaces. We set up the theoretical framework for its use in SMC with a focus on the ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Goldstein-Kac telegraph processes with random speeds: Path probabilities, likelihoods, and reported Lévy flights

Physical review. E, Statistical, nonlinear, and soft matter physics, 2015

The Goldstein-Kac telegraph process describes the one-dimensional motion of particles with consta... more The Goldstein-Kac telegraph process describes the one-dimensional motion of particles with constant speed undergoing random changes in direction. Despite its resemblance to numerous real-world phenomena, the singular nature of the resultant spatial distribution of each particle precludes the possibility of any a posteriori empirical validation of this random-walk model from data. Here we show that by simply allowing for random speeds, the ballistic terms are regularized and that the diffusion component can be well-approximated via the unscented transform. The result is a computationally efficient yet robust evaluation of the full particle path probabilities and, hence, the parameter likelihoods of this generalized telegraph process. We demonstrate how a population diffusing under such a model can lead to non-Gaussian asymptotic spatial distributions, thereby mimicking the behavior of an ensemble of Lévy walkers.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Handbook of Statistical Systems Biology

Systems biology is data-rich. Technological advances over the past 50 years have allowed us to pr... more Systems biology is data-rich. Technological advances over the past 50 years have allowed us to probe, map and interfere with biological organisms in a number of ways. Most notable are perhaps the sequencing efforts that are continuing to catalogue the genetic diversity of life, including our own species. A whole host of other techniques from biochemistry, molecular, cell and structural biology have been used to study the function of the protein products and other biomolecules that are encoded by these sequences. So we live in a ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Statistical analysis of network data and evolution on GPUs: High-performance statistical computing

Bookmarks Related papers MentionsView impact

Research paper thumbnail of A Network Analysis Primer

Statistical and Evolutionary Analysis of …, Jan 1, 2010

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Inference of Signalling Pathway Models

Handbook of Statistical Systems …, Jan 1, 2011

Skip to Main Content. ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Considerate approaches to constructing summary statistics for ABC model selection

Statistics and Computing, Jan 1, 2012

Abstract For nearly any challenging scientific problem evaluation of the likelihood is problemati... more Abstract For nearly any challenging scientific problem evaluation of the likelihood is problematic if not impossible. Approximate Bayesian computation (ABC) allows us to employ the whole Bayesian formalism to problems where we can use simulations from a model, but cannot evaluate the likelihood directly. When summary statistics of real and simulated data are compared—rather than the data directly—information is lost, unless the summary statistics are sufficient. Sufficient statistics are, however, not common but without them ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Variable Drug Response: Genetic Evaluation

eLS, Jan 1, 2006

Pharmacogenetics emerged in the late 1950s and early 1960s when variation in the metabolism of th... more Pharmacogenetics emerged in the late 1950s and early 1960s when variation in the metabolism of the antimalarial agents primaquine and isoniazid and the antihypertensive debrisoquine was shown to follow a simple Mendelian inheritance pattern. Following such discoveries, Motulsky set the conceptual foundation for pharmacogenetics by pointing out that common nonclinical or subclinical genetic variation may underlie genetic variation contributing to pharmacokinetic differences among individuals. The classical debrisoquine ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Biosimulation in Drug Development

Human Genomics, Jan 1, 2008

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Statistical and Evolutionary Analysis of Biological Networks

Page 1. Statistical and Evolutionary Analysis of Biological Networks imperial CoJJege Press Page ... more Page 1. Statistical and Evolutionary Analysis of Biological Networks imperial CoJJege Press Page 2. Statistical and Evolutionary Analysis of Biological Networks Page 3. This page intentionally left blank This page intentionally left blank Page 4. Statistical and Evolutionary Analysis of Biological Networks Imperial College Press ICP editors Michael PH Stumpf Imperial College London, UK Carsten Wiuf Aarhus University, Denmark Page 5. Published ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Statistical Null Models for Biological Network Analysis

Statistical and Evolutionary …, Jan 1, 2010

Statistical ensembles of random graphs serve as null models in the statistical analysis of real c... more Statistical ensembles of random graphs serve as null models in the statistical analysis of real complex networks. They encapsulate what are believed to be the generic properties of networks and describe the expected behaviour against which observed network data can be compared. Here we review the basic statistical physics underlying statistical ensembles of networks and show how we can exploit their properties. We also show how the simple statistical ensembles that have been used to describe networks can be improved by ...

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Language's place in nature

Trends in Ecology & Evolution, Jan 1, 2001

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Binomial subsampling

Proceedings of the Royal …, Jan 1, 2006

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Critical truths about power laws

Science, Jan 1, 2012

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Bayesian design strategies for synthetic biology

Interface Focus, Jan 1, 2011

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Inference of Temporally Varying Bayesian Networks

Bioinformatics (Oxford, England), Jan 1, 2012

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Parameter-free model discrimination criterion based on steady-state coplanarity

Proceedings of the National Academy of Sciences of the United States of America, Jan 1, 2012

We introduce a procedure for deciding when a mass-action model is incompatible with observed stea... more We introduce a procedure for deciding when a mass-action model is incompatible with observed steady-state data that does not require any parameter estimation. Thus, we avoid the difficulties of nonlinear optimization typically associated with methods based on parameter fitting. Instead, we borrow ideas from algebraic geometry to construct a transformation of the model variables such that any set of steady states of the model under that transformation lies on a common plane, irrespective of the values of the model parameters. Model rejection can then be performed by assessing the degree to which the transformed data deviate from coplanarity. We demonstrate our method by applying it to models of multisite phosphorylation and cell death signaling. Our framework offers a parameter-free perspective on the statistical model selection problem, which can complement conventional statistical methods in certain classes of problems where inference has to be based on steady-state data and the model structures allow for suitable algebraic relationships among the steady-state solutions.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Evolutionary characteristics of bacterial two-component systems

Advances in experimental medicine and biology, Jan 1, 2012

The evolution of biological systems is influenced by a number of factors and forces that have act... more The evolution of biological systems is influenced by a number of factors and forces that have acted in different combinations at different times to give rise to extant organisms. Here we illustrate some of the issues surrounding the data-driven evolutionary analysis of biological systems in the context of bacterial two-component systems (TCSs). TCSs are critical for bacteria to interact with their extracellular environment. A typical TCS consists of a histidine kinase on the membrane and a response regulator in the cytoplasm. Here we comprehensively characterise the extent to which these appear together across some 950 bacterial species and test for statistically significant patterns of correlated gain and loss. Our analysis provides evidence for correlated evolution but also a high level of evolutionary flexibility: at the sequence level, histidine kinases but especially response regulators belonging to different TCSs in a species show high levels of similarity, which may facilitate crosstalk as well as the recruitment of components into new compound signalling systems. We furthermore find that bacterial lifestyle has an overriding influence on the presence and absence of TCS; while in most TCSs either both or none of the two components are present, several TCSs tend to lose preferentially either the histidine kinase or response regulator component, which further supports the notion of reuse and reshuffling of these components in different TCS arrangements. We conclude by placing these findings in a wider context and discuss the implications for evolutionary systems biology more generally.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of Elucidating the in vivo phosphorylation dynamics of the ERK MAP kinase using quantitative proteomics data and Bayesian model selection

Molecular bioSystems, Jan 1, 2012

Ever since reversible protein phosphorylation was discovered, it has been clear that it plays a k... more Ever since reversible protein phosphorylation was discovered, it has been clear that it plays a key role in the regulation of cellular processes. Proteins often undergo double phosphorylation, which can occur through two possible mechanisms: distributive or processive. Which phosphorylation mechanism is chosen for a particular cellular regulation bears biological significance, and it is therefore in our interest to understand these mechanisms. In this paper we study dynamics of the MEK/ERK phosphorylation. We employ a model selection algorithm based on approximate Bayesian computation to elucidate phosphorylation dynamics from quantitative time course data obtained from PC12 cells in vivo. The algorithm infers the posterior distribution over four proposed models for phosphorylation and dephosphorylation dynamics, and this distribution indicates the amount of support given to each model. We evaluate the robustness of our inferential framework by systematically exploring different ways of parameterizing the models and for different prior specifications. The models with the highest inferred posterior probability are the two models employing distributive dephosphorylation, whereas we are unable to choose decisively between the processive and distributive phosphorylation mechanisms.

Bookmarks Related papers MentionsView impact

Research paper thumbnail of ABC-SysBio--approximate Bayesian computation in Python with GPU support

Bioinformatics (Oxford, England), Jan 1, 2010

Bookmarks Related papers MentionsView impact