Monday, 30th March
TimeTeller: a New Tool for Precision Circadian Medicine and Cancer Prognosis
David Rand, University of Warwick, UK
Recent research has shown that the circadian clock has a much more profound effect on human health than previously thought. I will present a machine-learning approach to measuring circadian clock functionality from the expression levels of key genes in a single tissue sample and then apply this to study survival in a breast cancer clinical trial.
A principal aim of circadian medicine is to develop techniques and methods to integrate the relevance of biological time into clinical practice. However, it is difficult to monitor the functional state of the circadian clock and its downstream targets in humans. Consequently, there is a critical need for tools to do this that are practical in a clinical context and our approach tackles this. We apply our algorithm to breast cancer and show that in a large cohort of patients with non-metastatic breast cancer the resulting dysfunction metric is a prognostic factor for survival providing evidence that it is independent of other known factors. While previous work in this area is focused on individual genes, our approach directly assesses the systemic functionality of a key regulatory system, the circadian clock, from one sample.
David Rand, University of Warwick, UK
Recent research has shown that the circadian clock has a much more profound effect on human health than previously thought. I will present a machine-learning approach to measuring circadian clock functionality from the expression levels of key genes in a single tissue sample and then apply this to study survival in a breast cancer clinical trial.
A principal aim of circadian medicine is to develop techniques and methods to integrate the relevance of biological time into clinical practice. However, it is difficult to monitor the functional state of the circadian clock and its downstream targets in humans. Consequently, there is a critical need for tools to do this that are practical in a clinical context and our approach tackles this. We apply our algorithm to breast cancer and show that in a large cohort of patients with non-metastatic breast cancer the resulting dysfunction metric is a prognostic factor for survival providing evidence that it is independent of other known factors. While previous work in this area is focused on individual genes, our approach directly assesses the systemic functionality of a key regulatory system, the circadian clock, from one sample.
Examining the collective dynamics and function of pancreatic islets of Langerhans with network science approaches
Marko Gosak, University of Maribor, Slovenia
The past two decades have witnessed the coming of age of network science as the central paradigm behind some of the most fascinating discoveries of the 21st century. The flexibility to define network nodes and edges to represent different aspects of real-life systems has also been employed to examine diverse biological settings at different scales. In the lecture I shall review the recent advances in the study of complex biological systems that were inspired and enabled by methods of network science. Specific focus will be given to the extraction of functional connectivity maps from high-resolution confocal calcium imaging in acute pancreas tissue slices to assess the collective behaviour of insulin secreting beta cell populations. Describing the intercellular connectivity patterns by means of network language has not only revealed that beta cell networks share many architectural similarities with several other real-life networks, such as small-worldness, heterogeneity, and modularity, but also leads to important new insights into the relationship between cellular metabolic activity and the orchestration of collective behaviour. Finally, I will concentrate on the emerging field of multilayer networks and highlight the potential offered by this framework in exploring the complex organization of tissues in both health and disease.
Marko Gosak, University of Maribor, Slovenia
The past two decades have witnessed the coming of age of network science as the central paradigm behind some of the most fascinating discoveries of the 21st century. The flexibility to define network nodes and edges to represent different aspects of real-life systems has also been employed to examine diverse biological settings at different scales. In the lecture I shall review the recent advances in the study of complex biological systems that were inspired and enabled by methods of network science. Specific focus will be given to the extraction of functional connectivity maps from high-resolution confocal calcium imaging in acute pancreas tissue slices to assess the collective behaviour of insulin secreting beta cell populations. Describing the intercellular connectivity patterns by means of network language has not only revealed that beta cell networks share many architectural similarities with several other real-life networks, such as small-worldness, heterogeneity, and modularity, but also leads to important new insights into the relationship between cellular metabolic activity and the orchestration of collective behaviour. Finally, I will concentrate on the emerging field of multilayer networks and highlight the potential offered by this framework in exploring the complex organization of tissues in both health and disease.
Temporal contact networks: data, data-driven models, sampling issues and prediction
Alain Barrat, CNRS and Aix-Marseille University, France
Face-to-face contacts between individuals play an important role in social interactions but determine also the potential transmission routes of infectious diseases, in particular of respiratory pathogens. An accurate description of these patterns is therefore of clear interest in epidemiology. While researchers have long relied on surveys and diaries to gather data concerning contacts between individuals, the availability of networked wearable devices is providing new ways to expose the interactions of individuals. In this talk, I will present results obtained by the SocioPatterns collaboration (www.sociopatterns.org), which focuses on measuring and modelling human contacts, and on using the gathered data and knowledge in particular for issues of relevance to the epidemiology of infectious diseases.
I will briefly describe the SocioPatterns sensing platform and some of the datasets collected in the last 10 years, and present an example of how such data can be used to investigate the relative efficiency of targeted mitigation measures such as the targeted closure of classes in schools in case of a disease outbreak. I will then discuss two issues arising when data is incomplete. First, numerical simulations of a disease model using incomplete contact network data can lead to incorrect predictions and in particular to underestimate the epidemic risk. I will show how it is possible to compensate for such biases by building surrogate data sets using the statistical properties of the incomplete data. Second, I will discuss the case in which the contact network is known and the state of only a fraction of the nodes at random times can be observed during a single outbreak: I will show how, using recently developed node embedding methods, it is possible to predict the dynamical state of all nodes at all times, even without detailed knowledge of the nature of the spreading process. This makes it possible to estimate the temporal evolution of the entire system, with potential impact for nowcasting infectious disease dynamics.
Alain Barrat, CNRS and Aix-Marseille University, France
Face-to-face contacts between individuals play an important role in social interactions but determine also the potential transmission routes of infectious diseases, in particular of respiratory pathogens. An accurate description of these patterns is therefore of clear interest in epidemiology. While researchers have long relied on surveys and diaries to gather data concerning contacts between individuals, the availability of networked wearable devices is providing new ways to expose the interactions of individuals. In this talk, I will present results obtained by the SocioPatterns collaboration (www.sociopatterns.org), which focuses on measuring and modelling human contacts, and on using the gathered data and knowledge in particular for issues of relevance to the epidemiology of infectious diseases.
I will briefly describe the SocioPatterns sensing platform and some of the datasets collected in the last 10 years, and present an example of how such data can be used to investigate the relative efficiency of targeted mitigation measures such as the targeted closure of classes in schools in case of a disease outbreak. I will then discuss two issues arising when data is incomplete. First, numerical simulations of a disease model using incomplete contact network data can lead to incorrect predictions and in particular to underestimate the epidemic risk. I will show how it is possible to compensate for such biases by building surrogate data sets using the statistical properties of the incomplete data. Second, I will discuss the case in which the contact network is known and the state of only a fraction of the nodes at random times can be observed during a single outbreak: I will show how, using recently developed node embedding methods, it is possible to predict the dynamical state of all nodes at all times, even without detailed knowledge of the nature of the spreading process. This makes it possible to estimate the temporal evolution of the entire system, with potential impact for nowcasting infectious disease dynamics.
Consequences of delays and imperfect implementation of isolation in epidemic control
Serhiy Yanchuk, Technische Universität Berlin, Germany
Abstract: Isolation has been the main control strategy of unforeseen epidemic outbreaks. When implemented in full and without delay, isolation is very effective. However, flawless implementation is seldom feasible in practice. We present an epidemic model called SIQ with an isolation protocol, focusing on the consequences of delays and incomplete identification of infected hosts. The continuum limit of this model is a system of Delay Differential Equations, the analysis of which reveals the dependence of epidemic evolution on model parameters including disease reproductive number, isolation probability, speed of identification of infected hosts and recovery rates. Our model offers estimates on minimum response capabilities needed to curb outbreaks, and predictions of endemic states when containment fails. Critical response capability is expressed explicitly in terms of parameters that are easy to obtain, to assist in the evaluation of funding priorities involving preparedness and epidemics management.
This is a joint work with L.-S. Young (Courant Institute of Mathematical Sciences, New York), S. Ruschel (University of Auckland), and T. Pereira (University of Sao-Paolo and Imperial College London, UK)
Serhiy Yanchuk, Technische Universität Berlin, Germany
Abstract: Isolation has been the main control strategy of unforeseen epidemic outbreaks. When implemented in full and without delay, isolation is very effective. However, flawless implementation is seldom feasible in practice. We present an epidemic model called SIQ with an isolation protocol, focusing on the consequences of delays and incomplete identification of infected hosts. The continuum limit of this model is a system of Delay Differential Equations, the analysis of which reveals the dependence of epidemic evolution on model parameters including disease reproductive number, isolation probability, speed of identification of infected hosts and recovery rates. Our model offers estimates on minimum response capabilities needed to curb outbreaks, and predictions of endemic states when containment fails. Critical response capability is expressed explicitly in terms of parameters that are easy to obtain, to assist in the evaluation of funding priorities involving preparedness and epidemics management.
This is a joint work with L.-S. Young (Courant Institute of Mathematical Sciences, New York), S. Ruschel (University of Auckland), and T. Pereira (University of Sao-Paolo and Imperial College London, UK)
A Method for Determining the Likely Next Destinations of COVID-19
Julien Arino, University of Manitoba, Canada
Julien Arino, University of Manitoba, Canada
Tuesday, 31st March
Epidemic Spreading on Temporal Networks: a Contact-Based Modeling Framework
Philipp Hövel, University College Cork, Ireland
Many networks exhibit time-dependent topologies with edges existing for some time or weights subject to temporal fluctuations. This is particularly important, if the evolution of the network topology acts on a timescale similar to the local node dynamics [1]. The availability of data with high temporal resolution, which can be mathematically described as temporal networks, allows to improve, for instance, epidemiological models and estimate the impact of potential outbreaks from past contact data.
In this presentation, we present a contact-based model to study the spreading of epidemics by means of extending the dynamic message-passing approach by B. Karrer and M. Newman [2] to temporal networks. We shift the perspective from node- to edge-centric quantities [3]. This enables accurate modeling of Markovian susceptible-infected-recovered outbreaks on time-varying trees, i.e., temporal networks with a loop-free underlying topology. On arbitrary graphs, the proposed contact-based (CB) model incorporates potential structural and temporal heterogeneities of the contact network and improves analytic estimations with respect to the node-centric, individual-based (IB) approach at a low computational and conceptual cost. The method is based on the largest eigenvalue of the infection propagator matrix, which determines the disease propagation in the low prevalence limit and takes into account the full temporal and topological information up to the observation time without relying on extensive Monte-Carlo simulations and a subsequent parameter fit.
Within the proposed framework, we derive an analytical expression for the epidemic threshold on temporal networks and demonstrate the feasibility of this method on empirical data (face-to-face interaction network at a conference and livestock-trade network). As a potential application to public health institutions, we provide risk estimates for global outbreaks and analyze their temporal and spatial variability in terms of the critical infection probability. Our quantitative comparison demonstrates that spectral methods provide a lower bound with varying degree of accuracy depending on the network details. Despite their heterogeneity in size and activity, we find for all networks that the CB model outperforms the IB approach.
References:
[1] M. Konschake, H. H. K. Lentz, F. Conraths, P. Hövel, and T. Selhorst: On the Robustness of In- and Out-Components in a Temporal Graph, PLoS ONE 8, e55223 (2013).
[2] B. Karrer and M. E. J. Newman: Message passing approach for general epidemic models, Phys. Rev. E 82, 016101 (2010).
[3] A. Koher, H.H.K. Lentz, J. P Gleeson, and P. Hövel: Contact-Based Model for Epidemic Spreading on Temporal Networks, Phys. Rev. X 9, 031017 (2019).
Philipp Hövel, University College Cork, Ireland
Many networks exhibit time-dependent topologies with edges existing for some time or weights subject to temporal fluctuations. This is particularly important, if the evolution of the network topology acts on a timescale similar to the local node dynamics [1]. The availability of data with high temporal resolution, which can be mathematically described as temporal networks, allows to improve, for instance, epidemiological models and estimate the impact of potential outbreaks from past contact data.
In this presentation, we present a contact-based model to study the spreading of epidemics by means of extending the dynamic message-passing approach by B. Karrer and M. Newman [2] to temporal networks. We shift the perspective from node- to edge-centric quantities [3]. This enables accurate modeling of Markovian susceptible-infected-recovered outbreaks on time-varying trees, i.e., temporal networks with a loop-free underlying topology. On arbitrary graphs, the proposed contact-based (CB) model incorporates potential structural and temporal heterogeneities of the contact network and improves analytic estimations with respect to the node-centric, individual-based (IB) approach at a low computational and conceptual cost. The method is based on the largest eigenvalue of the infection propagator matrix, which determines the disease propagation in the low prevalence limit and takes into account the full temporal and topological information up to the observation time without relying on extensive Monte-Carlo simulations and a subsequent parameter fit.
Within the proposed framework, we derive an analytical expression for the epidemic threshold on temporal networks and demonstrate the feasibility of this method on empirical data (face-to-face interaction network at a conference and livestock-trade network). As a potential application to public health institutions, we provide risk estimates for global outbreaks and analyze their temporal and spatial variability in terms of the critical infection probability. Our quantitative comparison demonstrates that spectral methods provide a lower bound with varying degree of accuracy depending on the network details. Despite their heterogeneity in size and activity, we find for all networks that the CB model outperforms the IB approach.
References:
[1] M. Konschake, H. H. K. Lentz, F. Conraths, P. Hövel, and T. Selhorst: On the Robustness of In- and Out-Components in a Temporal Graph, PLoS ONE 8, e55223 (2013).
[2] B. Karrer and M. E. J. Newman: Message passing approach for general epidemic models, Phys. Rev. E 82, 016101 (2010).
[3] A. Koher, H.H.K. Lentz, J. P Gleeson, and P. Hövel: Contact-Based Model for Epidemic Spreading on Temporal Networks, Phys. Rev. X 9, 031017 (2019).
(Un-)supervised Learning of Cell Population Structure and Cell State Transitions from Single-Cell Snapshot Data
Manfred Claassen, ETH Zürich, Switzerland
Single cell technologies are transforming the investigation of cell states in biological processes and heterogeneous tissues. Identification of cell types/states and reconstruction of cell state sequences from such data typically rely on unsupervised approaches, and therefore fail to take advantage of phenotype information coming along with single experiments (e.g. disease status or time points), and in presence of confounding variation fail to identify phenotype-associated cell subsets. To fill this gap, we have developed a suite of supervised learning approaches to identify phenotype associated cell subsets from high-dimensional single cell data. Specifically, we demonstrate how (deep and shallow) convolutional neural networks can identify of rare CMV infection and multiple sclerosis-associated cell subsets in peripheral blood, and extremely rare leukemic blast populations in minimal residual disease-like situations, as well as identification of morphological patterns associated with severity of prostate cancer. Further we recently developed psupertime, a supervised approach to pseudotime ordering. We demonstrate superior ordering for five single cell RNA-seq studies to conventional unsupervised pseudotime ordering techniques. We expect these supervised learning approaches to tap the potential of multi-experiment studies to come by enabling the identification and molecular characterization of phenotype-associated cell subpopulations in the complex tissue context across health and disease.
Manfred Claassen, ETH Zürich, Switzerland
Single cell technologies are transforming the investigation of cell states in biological processes and heterogeneous tissues. Identification of cell types/states and reconstruction of cell state sequences from such data typically rely on unsupervised approaches, and therefore fail to take advantage of phenotype information coming along with single experiments (e.g. disease status or time points), and in presence of confounding variation fail to identify phenotype-associated cell subsets. To fill this gap, we have developed a suite of supervised learning approaches to identify phenotype associated cell subsets from high-dimensional single cell data. Specifically, we demonstrate how (deep and shallow) convolutional neural networks can identify of rare CMV infection and multiple sclerosis-associated cell subsets in peripheral blood, and extremely rare leukemic blast populations in minimal residual disease-like situations, as well as identification of morphological patterns associated with severity of prostate cancer. Further we recently developed psupertime, a supervised approach to pseudotime ordering. We demonstrate superior ordering for five single cell RNA-seq studies to conventional unsupervised pseudotime ordering techniques. We expect these supervised learning approaches to tap the potential of multi-experiment studies to come by enabling the identification and molecular characterization of phenotype-associated cell subpopulations in the complex tissue context across health and disease.
Protein interaction networks for single-cell RNA-seq analysis
Florian Klimm, Imperial College London/ University of Cambridge, UK
Protein–protein interaction networks (PPINs) describe molecular reactions between proteins and thus can be used to predict biological functions of proteins. Protein interaction data, however, lacks biological context because gene expression is cell-type dependent. Recent advances in single-cell RNA sequencing (scRNA-seq) have allowed researchers to explore gene expression at a cellular level. In this talk, we will discuss scPPIN, a method for integrating single-cell RNA sequencing data with protein–protein interaction networks to detect active modules in cells of different transcriptional states. We achieve this by clustering RNA-sequencing data, identifying differentially expressed genes, constructing node- weighted PPINs, and finding the maximum-weight connected subgraphs with an exact Steiner-tree approach. As a case study, we investigate RNA-sequencing data from human liver spheroids but the techniques described here are applicable to other organisms and tissues. The scPPIN method allows us to expand the output of differential expressed genes analysis with information from protein interactions. We find that different transcriptional states have different subnetworks of the PPIN significantly enriched which represent biological pathways. In these pathways, scPPIN also identifies proteins that are not differentially expressed but of crucial biological function (e.g., as receptors) and therefore reveals biology beyond a standard differentially expressed gene analysis.
Florian Klimm, Imperial College London/ University of Cambridge, UK
Protein–protein interaction networks (PPINs) describe molecular reactions between proteins and thus can be used to predict biological functions of proteins. Protein interaction data, however, lacks biological context because gene expression is cell-type dependent. Recent advances in single-cell RNA sequencing (scRNA-seq) have allowed researchers to explore gene expression at a cellular level. In this talk, we will discuss scPPIN, a method for integrating single-cell RNA sequencing data with protein–protein interaction networks to detect active modules in cells of different transcriptional states. We achieve this by clustering RNA-sequencing data, identifying differentially expressed genes, constructing node- weighted PPINs, and finding the maximum-weight connected subgraphs with an exact Steiner-tree approach. As a case study, we investigate RNA-sequencing data from human liver spheroids but the techniques described here are applicable to other organisms and tissues. The scPPIN method allows us to expand the output of differential expressed genes analysis with information from protein interactions. We find that different transcriptional states have different subnetworks of the PPIN significantly enriched which represent biological pathways. In these pathways, scPPIN also identifies proteins that are not differentially expressed but of crucial biological function (e.g., as receptors) and therefore reveals biology beyond a standard differentially expressed gene analysis.
Optimising for infectious disease control using algorithmic graph theory
Jessica Enright, University of Glasgow, UK
Algorithmic graph theory is awash with tools for solving combinatorial optimisation problems on graphs (or networks, if you prefer), particularly graph modification problems. I will discuss several pieces of work that aim to use these optimisation tools to solve problems relevant to disease control (mainly in livestock, though I shall try to make these relevant to humans as well), including scheduling of temporal networks to minimize infectious risk, optimal vaccination defense policies, and geographic partitioning
Jessica Enright, University of Glasgow, UK
Algorithmic graph theory is awash with tools for solving combinatorial optimisation problems on graphs (or networks, if you prefer), particularly graph modification problems. I will discuss several pieces of work that aim to use these optimisation tools to solve problems relevant to disease control (mainly in livestock, though I shall try to make these relevant to humans as well), including scheduling of temporal networks to minimize infectious risk, optimal vaccination defense policies, and geographic partitioning
Recurrence Quantification Analysis of Dynamic Brain Networks and Computational Models to Inform Presurgical Evaluation of Epilepsy
Marinho Lopes, University of Bristol, UK
The brain is a complex network whose function results from multiscale spatio-temporal dynamics. Traditional approaches to studying brain function (and dysfunction) have primarily focused on the structure of static brain networks. However, functional brain networks are dynamic because they depend on time-evolving brain activity. In this talk I will present a new framework to assess the dynamics of brain networks based on recurrence analysis. The framework uses recurrence plots and recurrence quantification analysis to characterize dynamic networks. For resting-state magnetoencephalographic dynamic functional networks (dFNs), we found that functional networks recur more quickly in people with epilepsy than healthy controls. This suggests that recurrence of dFNs may be used as a biomarker of epilepsy. For stereo electroencephalography data, we found that dFNs involved in epileptic seizures emerge before seizure onset, and recurrence analysis allows us to detect seizures. We further observed distinct dFNs before and after seizures, which may inform neurostimulation strategies to prevent seizures [1].
I will also present a computational modelling framework to assess epilepsy lateralisation from scalp EEG [2]. Epilepsy lateralisation is paramount during presurgical assessment of people with drug-resistant epilepsy. We use eLORETA to map source activities from seizure epochs recorded from scalp EEG and obtain functional networks using the phase-locking value. The networks are then studied using a mathematical model of epilepsy. By removing different regions of interest from the network and simulating their impact on the network’s ability to generate seizures in silico, the framework provides predictions of epilepsy lateralization. The framework proved useful in assessing epilepsy lateralization in 12 out of 15 individuals considered. These results show promise for the use of this framework to better interrogate scalp EEG and aid clinicians in presurgical assessment of people with epilepsy.
[1] https://arxiv.org/abs/2001.03761
[2] Lopes, M. A., et al. "Computational modelling in source space from scalp EEG to inform presurgical evaluation of epilepsy." Clinical Neurophysiology 131.1 (2020): 225-234.
Marinho Lopes, University of Bristol, UK
The brain is a complex network whose function results from multiscale spatio-temporal dynamics. Traditional approaches to studying brain function (and dysfunction) have primarily focused on the structure of static brain networks. However, functional brain networks are dynamic because they depend on time-evolving brain activity. In this talk I will present a new framework to assess the dynamics of brain networks based on recurrence analysis. The framework uses recurrence plots and recurrence quantification analysis to characterize dynamic networks. For resting-state magnetoencephalographic dynamic functional networks (dFNs), we found that functional networks recur more quickly in people with epilepsy than healthy controls. This suggests that recurrence of dFNs may be used as a biomarker of epilepsy. For stereo electroencephalography data, we found that dFNs involved in epileptic seizures emerge before seizure onset, and recurrence analysis allows us to detect seizures. We further observed distinct dFNs before and after seizures, which may inform neurostimulation strategies to prevent seizures [1].
I will also present a computational modelling framework to assess epilepsy lateralisation from scalp EEG [2]. Epilepsy lateralisation is paramount during presurgical assessment of people with drug-resistant epilepsy. We use eLORETA to map source activities from seizure epochs recorded from scalp EEG and obtain functional networks using the phase-locking value. The networks are then studied using a mathematical model of epilepsy. By removing different regions of interest from the network and simulating their impact on the network’s ability to generate seizures in silico, the framework provides predictions of epilepsy lateralization. The framework proved useful in assessing epilepsy lateralization in 12 out of 15 individuals considered. These results show promise for the use of this framework to better interrogate scalp EEG and aid clinicians in presurgical assessment of people with epilepsy.
[1] https://arxiv.org/abs/2001.03761
[2] Lopes, M. A., et al. "Computational modelling in source space from scalp EEG to inform presurgical evaluation of epilepsy." Clinical Neurophysiology 131.1 (2020): 225-234.