Journées annuelles du groupe de travail (4ème édition)
Date : Lundi 2 (toute la journée) et mardi 3 juillet (matin) 2018
Lieu : site Saint-Charles de l'université d'Aix-Marseille à Marseille.
Organisateurs : Elisabeth Remy, Grégory Batt, Cédric Lhoussaine et Anne Siegel
La quatrième édition des journées annuelles du GT Bioss va se dérouler juste avant les Journées Ouvertes Biologie, Informatique et Mathématiques (JOBIM).
L'inscription, gratuite mais obligatoire, se fait en remplissant le formulaire accessible ici.
Olivier Bernard, équipe projet BioCore, Inria.
François Fages, équipe projet Lifeware, Inria.
Bertie Gottgens, Cambridge Institute for Medical Research.
Heike Siebert, DFG-Research Center Matheon, Berlin.
Lundi 2 Juillet
10h00-10h45 - Conférencière invitée - Heike Siebert (Université de Berlin)- A Boolean Look at Synthetic Biology - Finding Cell Classifiers Using Answer Set Programming
10h45-11h05 - Adrien Richard (I3S, Nice) - Fixing monotone boolean networks asynchronously
11h30-11h50 - Stephanie Chevalier (LRI) - A logical approach to identify Boolean Networks that model cell differentiation
11h55-12h15 - Maxime Folschette (Irset/Irisa) - GULA: Semantics-Free Learning of a Biological Regulatory Networks from a Synchronous, Asynchronous or Generalized State Graph
12h20-12h40 - Aurelien Naldi (ENS Paris) - Similarities and complementarity of positive feedback circuits and stable motifs in logical regulatory networks
12h40- 14h00 - Pause déjeuner
14h00-14h45 - Conférencier invité -Bertie Göttgens (Cambridge institute for Medical Research) - Reconstructing Cell States, Lineage trajectories and Regulatory Networks from Single cells Molecular profiles.
14h50-15h10 - Alberto Valdeolivas (I2M) - A Multiplex Network approach to Premature Aging Diseases
15h30-15h50 - Céline Hernandez (ENS Paris) - Dynamical modelling of T cell co-inhibitory pathways to predict anti-tumour responses to checkpoint inhibitors
15h50-16h10 - Eugenia Oshurko (LIP) - Representation and aggregation of cellular signalling knowledge in KAMI
16h15-16h35 - Sébastien Légaré (LIP) - Biocuration and rule-based modelling of protein interaction networks in KAMI
17h00-17h45 - Conférencier invité - François Fages (Inria Saclay) - Computer-aided biochemical programming of synthetic micro reactors as diagnostic devices
17h50-18h10 - Loic Pauleve (LRI) - The CoLoMoTo Interactive Notebook: Accessible and Reproducible Computational Analyses for Qualitative Biological Networks 18h15-18h35 - Discussion Bioss.
Mardi 3 Juillet
09h00-09h45 - Conférencier invité - Olivier Bernard (Inria Nice Sophia Antipolis) - Dynamical Reduction of Metabolic Networks. Application to Microalgae
09h50-10h10 - Nils Giordano (LS2N) - Using co-activity networks to reveal the structure of planktonic symbioses in the global ocean
10h30-10h50 - Ghuvan Grimaud (Biomathematica) - EvoDRUM: an evolutionary systems biology framework to investigate the origin of early metabolisms
10h55-11h15 - Anne Siegel (IRISA) - Learning boolean rules for the regulatory control of metabolism: a case study
11h20-11h40 Thibault Etienne (Ibis, Inria) - Coordination of mRNA stability and cell physiology in bacteria: a modelling study
Stephanie Chevalier (LRI) - A logical approach to identify Boolean Networks that model cell differentiation.
Thibault Etienne (Ibis, Inria) - Coordination of mRNA stability and cell physiology in bacteria: a modelling study
Thibault Etienne, Laurence Girbal, Muriel Cocaign-Bousquet, Delphine Ropers
The adaptation of bacterial physiology to environmental fluctuations involves system-wide changes of metabolism and gene expression. This reprogramming of the cell takes place at two different levels: on a global scale through the adjustment of the level and activity of components of the gene expression machinery (ribosomes...), and locally, through the adjustment of the concentration of regulators specifically coordinating the cell response to the new environmental conditions.
The different regulatory levels are interlaced and form large biochemical networks, whose dynamic functioning is not intuitive. Among these regulatory levels, recent studies have shown that post-transcriptional regulations are more important than usually assumed. Contrary to the often-made assumption in bacteria, protein and mRNA levels are not proportional and mRNA stability (typically a few minutes) varies with the translational activity, the cell growth rate and the concentration of regulators (small RNAs, HFQ,...). How these interlocked control mechanisms adjust mRNA half-life to cell physiology remains largely unknown. In our study, we tackle this question by means of mathematical modelling using available times-series -omics data in Escherichia coli (transcriptomics and stabilomics). Our objective is to provide a mechanistic explanation of mRNA degradation profiles obtained at various growth rates.
We develop a structural model of mRNA degradation in E.coli based on Michaelis-Menten kinetics. In this model, the individual parameters vary with the nature of the mRNA and the cell growth rate. A mixed-effect modelling framework is used to take into account the variability of these parameters: using the genome-wide - omics data, we estimate the mean parameters describing the population of mRNAs and the variance parameters, which allow to reproduce the degradation profile of each mRNA in each condition. The analysis of mean parameter values informs us on the global regulatory effects, while the parameter variances reflect the specific regulatory mechanisms.
Maxime Folschette (Irset/Irisa) - GULA: Semantics-Free Learning of a Biological Regulatory Networks from a Synchronous, Asynchronous or Generalized State Graph
The automatic learning of an interaction graph from the sole observation of its dynamics is an ongoing challenge. An example is the existing LFIT algorithm which learns and refines logic rules representing a model, from a set of state transitions representing its dynamics. Starting from the learning of purely deterministic synchronous Boolean systems, several versions have been developed in order to tackle dynamics with memory, inconsistencies or with multi-valued variables. However, all of them rely on the knowledge of the underlying semantics, that is, the update scheme of the variables. This work intents to free the learning process from this knowledge.
With GULA (General Usage LFIT Algorithm), we focus on three different semantics: synchronous (all variables must update their value between two discrete time steps), asynchronous (exactly one variable must do so) or general (any subset of variables may do so). The learning presented here is based on the refinement of logic rules (a set of conditional atoms and a conclusion atom) that represent the possibility for a variable to change its value under some conditions on the current state. Such rules only represent the potentiality of a change, which makes them independent of the semantics. Nevertheless, we also exhibit some properties that characterize the dynamics of the three given semantics, allowing to correctly interpret the rules of the final regulatory graph.
This work opens many outcomes. The most pressing is finding a broader characterization of what a “learnable” semantics is, allowing to generalize the scope of this approach. Furthermore, the semantics itself could be learned along with the rules, allowing to entirely learn a system. Finally, getting rid of the arbitrary but mandatory discretization step would allow to directly learn from the gene expression measurements, as already proposed with ACEDIA.
Ghuvan Grimaud (Biomathematica) - EvoDRUM: an evolutionary systems biology framework to investigate the origin of early metabolisms
Ghjuvan Grimaud, Elena Litchman, Christopher Klausmeier
The origin of the fundamental metabolic pathways and the subsequent rise of the great metabolic diversity of microbes are two major steps in life’s evolution on Earth and potentially other habitable planets. Understanding how different metabolisms may arise, what conditions select for different types of metabolic networks, and how they assemble to form ecological communities are key questions for the origin of life. The evolutionary emergence of diverse metabolisms depends not only on environmental conditions but also on microbial interactions such as competition and mutualism. Ecological interactions are being increasingly recognized as a driving force of evolutionary diversification in different groups of organisms, including microbes . So far, the role of microbial interactions in the origin of metabolic pathways under dynamic conditions has not been investigated in detail. Here we propose to combine two novel modeling approaches from two disparate disciplines (systems biology and evolutionary ecology) to explore how microbial metabolic networks arise and evolve in dynamic community contexts. We embed a recently developed metabolic modeling approach for the elementary flux mode analysis under nonequilibrium conditions (the Dynamic Reduction of Unbalanced Metabolism, DRUM) in an eco-evolutionary modeling framework of trait evolution (Adaptive Dynamics [3,4]) to investigate how different metabolic networks arise and compete in different environments. The resulting new Evolutionary Systems Biology mathematical framework (evoDRUM) is a powerful tool that allows extensive explorations of how early metabolisms appeared and were maintained by natural selection and, thus, is useful for the field of early microbial evolution. EvoDRUM extends and modifies the idea of gathering the evolutionarily possible reactions by defining a large — ideally universal — mutation space in which evolution can proceed. In line with the Adaptive Dynamics framework, evolution occurs by a step-by-step mutant/resident invasion dynamics, with a defined mutation rate. The novelty of the proposed approach is that it investigates the metabolically explicit trait changes and evolution as a result of selection through competitive interactions of different phenotypes, and allows the incorporation of metabolite accumulation and evolutionary innovations. First applied to simple metabolic networks with several resources and temporally fluctuating conditions, we then use it for genome-scale metabolic networks.
1. Brodie, J., Ball, S.G., Bouget, F.-Y., Chan, C.X., De Clerck, O., Cock, J.M., Gachon, C., Grossman, A.R., Mock, T., Raven, J.A., Saha, M., Smith, A.G., Vardi, A., Yoon, H.S., and Bhattacharya, D. (2017). Biotic interactions as drivers of algal origin and evolution. New Phytologist 216, 670-681.
2. Baroukh C., Munoz-Tamayo R., Steyer J.P. and Bernard O. (2014). DRUM: A new framework for metabolic modeling under non- balanced growth. Application to the carbon metabolism of unicellular microalgae. PloS one, 9 (8), e104499.
3. Dieckmann U. and Law R. (1996). The dynamical theory of coevolution: A derivation from stochastic ecological processes. Journal of Mathematical Biology, 34, 579-612.
5. Geritz S., Kisdi E., Meszéna G. and Metz J. (1998). Evolutionarily singular strategies and the adaptive growth and branching of the evolutionary tree. Evolutionary Ecology, 12, 35-57.
5. Szappanos B., Fritzemeier J., Csörgo B., Lazar V., Lu X., Fekete G., Balint B., Herczeg R., Nagy I., Notebaart R.A. et al. (2016). Adaptive evolution of complex innovations through stepwise metabolic niche expansion. Nature communications, 7.
Céline Hernandez (ENS Paris) - Dynamical modelling of T cell co-inhibitory pathways to predict anti-tumour responses to checkpoint inhibitors
Céline Hernandez(1), Aurélien Naldi(1), Wassim Abou-Jaoudé(1), Guillaume Voisinne(2), Romain Roncagalli(2), Bernard Malissen(2), Morgane Thomas-Chollier(1), Denis Thieffry(1)
(1) Computational Systems Biology team, Institut de Biologie de l’Ecole Normale Supérieure (IBENS), CNRS UMR8197, INSERM U1024, Ecole Normale Supérieure, PSL Université, 75005 Paris, France
(2) Centre d'Immunologie de Marseille-Luminy, Aix Marseille Université, INSERM U1104, CNRS UMR7280, 13288 Marseille, France
In recent years, it has been recognised that T cells often display a reduced ability to eliminate cancer cells and that expression of co-inhibitors at their surface accounts for their compromised function. Antibodies blocking the functions of these co-inhibitors (checkpoint inhibitors) have become standard treatment for metastatic melanoma , leading to a revival in the study of T cell co-inhibitors. However, our understanding of the immunobiology of T cell co-inhibitors and of their harmful role during anti-tumour responses remains fragmentary. Despite some biochemical studies, a mechanistic understanding at the system-level of the modulation of T cell function by co-inhibitors has remained elusive.
To overcome these limitations, we aim at delineating the mechanisms through which co-inhibitory molecules, such as PD-1 and CTLA-4, impede T cell functions at the system-level. To reach this goal, we use computational methods to map and model TCR co-signalling pathways, and ultimately predict cell responses to perturbations.
First, we focused on the development of comprehensive annotated molecular maps (using the software CellDesigner ) based on the curation of scientific literature, in parallel with automated queries to public databases and protein-protein graph reconstruction. Next, using the software GINsim , these maps and protein networks are translated into a regulatory graph integrating current knowledge. The challenge is then to properly model concurrent intracellular processes, along with feedback control mechanisms. To cope with this complexity, we explored some network modules using a Rule-based formalism , in order to evaluate concurrent biological hypotheses and help specify logical rules recapitulating observed component behaviour back into the logical model. This model will be used to predict cell response to single or multiple perturbations, and thereby pave the way to the delineation of novel experiments, which will in turn be used to refine the maps and model.
This integrated system-level view of the mechanisms of action of key T cell co-inhibitors in cancer will further provide a rationale for designing and evaluating drugs targeting T cell co-inhibitory pathways in anti-cancer immunotherapy.
1. Simpson TR, Li F, Montalvo-Ortiz W, Sepulveda MA, Bergerhoff K, Arce F, Roddie C, Henry JY, Yagita H, Wolchok JD, Peggs KS, Ravetch JV, Allison JP, Quezada SA (2013). Fc-dependent depletion of tumor-infiltrating regulatory T cells co-defines the efficacy of anti-CTLA-4 therapy against melanoma. The Journal of experimental medicine 210(9): 1695–710.
4. Feret J, Danos V, Krivine J, Harmer R, Fontana W (2009). Internal coarse-graining of molecular systems. Proceedings of the National Academy of Sciences of the USA 106(16): 6453-8
Sébastien Légaré (LIP) - Biocuration and rule-based modelling of protein interaction networks in KAMI.
KAMI, the Knowledge Aggregator and Model Instantiator, is a software for biocuration and modelling of molecular interaction networks. It provides a knowledge representation to unambiguously express the details of biomolecular interactions. This representation can be built either programmatically or graphically via the KamiStudio interface. To assist users in curating their biological knowledge, KAMI is organised in two distinct layers: a network and a set of individual interactions called nuggets. Once a new nugget is built, it can be automatically aggregated to the network. The software then performs a series of tests to ensure consistency including duplicate search, biological database grounding and semantic checking. This greatly facilitates biocuration as users do not need to have the complete network in mind to add new data. Furthermore, interaction networks represented in KAMI can be directly converted to rule-based models in the Kappa language for simulation and analysis. In this talk, we will present the use of KAMI through a model of tyrosine phosphorylation involved in cell signaling. This example is well suited to showcase the advantages of the rule-based strategy. In particular, we will demonstrate the use of causality analysis to discover pathways in the model that were not explicitly input by the user.
Aurélien Naldi, ENS Paris - Similarities and complementarity of positive feedback circuits and stable motifs in logical regulatory networks
Aurélien Naldi, Denis Thieffry
Discrete qualitative models have been widely used to study complex biological regulatory networks. The increasing complexity of the systems of interest calls for efficient analysis methods, and in particular approaches directly relating the structure of the network to its dynamical properties.
The study of feedback circuits, based on the seminal work of R. Thomas, is a prominent example of such approaches: positive circuits are associated to the co-existence of multiple attractors, while negative circuits are associated to sustained oscillations . The properties of isolated circuits and of some simple combinations of circuits have been formally characterised , however their precise roles once embedded in complex networks remain unknown. An embedded circuit is called “functional” when the values of its regulators allow it to behave as an isolated circuit.
Stable motifs (also called symbolic steady states) have recently been proposed to efficiently identify attractors of such models [3,4]. Each stable motif represent a partial assignment of model components such that all successors of the matching states also belong to the motif.
The identification of stable motifs was recently added to the GINsim software , which already supported the identification of functional circuits. Based on the availability of these two analysis methods in the same software tool, we further explore the connection between the classical feedback circuits and stable motifs. The core of each stable motif is formed by a (group of) positive circuits settled in one of their two stable configurations. The resulting stability is often associated to functional positive circuits which can sustain their own functionality contexts. Stable motifs can further arise from non-functional positive circuits, which can be locked in only one of their two stable configurations.
 Comet et al. (2013). On circuit functionality in boolean networks. Bulletin of Mathematical Biology 75: 906-19.
 Remy et al. (2016). Boolean Dynamics of Compound Regulatory circuits. In : Rogato A, Zazzu V, Guarracino MR (Eds.). Dynamics of Mathematical Models in Biology. Springer International Publishing, pp. 43-53.
 Zañudo & Albert (2013). An effective network reduction approach to find the dynamical repertoire of discrete dynamic networks. Chaos 23: 025111.
 Klarner et al. (2014). Computing Symbolic Steady States of Boolean Networks. Lecture Notes in Computer Sciences 8751: 561-70.  http://ginsim.org
Eugenia Oshurko (LIP) - Representation and aggregation of cellular signalling knowledge in KAMI
Rule-based modelling has proven to be a successful approach for study- ing complex systems of cellular signalling. A rule-based language Kappa has been actively developed and used in recent years. However, building and curating big explanatory models using Kappa rules is challenging and cumbersome. To tackle exactly this problem we propose a bio-curation tool called KAMI (Knowledge Aggregator and Model Instantiator), which allows gradual semi-automatic aggregation of PPIs of different provenance, their annotation, visualisation and further instantiation to concrete rule-based models (including automatic generation of Kappa rules).
Models in KAMI are accommodated using a specially designed graph- based knowledge representation system which provides robust mechanisms for incremental aggregation of partial knowledge, its audit, update, and transfer to various representations. In this talk we will present this knowl- edge representation system, its properties and the mechanisms for knowledge update based on graph rewriting. We will also focus on its instance used in KAMI to represent models of cellular signalling systems. Then we will speak about the strategy of automatic knowledge aggregation that exploits the properties of this system. And finally, we will show how KAMI uses domain-specific background knowledge (e.g. semantics of conserved pro- tein domains, definitions of protein families, splice variants and mutants) to sharpen aggregated models.
Loic Pauleve (LRI) - The CoLoMoTo Interactive Notebook: Accessible and Reproducible Computational Analyses for Qualitative Biological Networks
Joint work with A Naldi, C Hernandez, N Levy, G Stoll, P Monteiro, C Chaouiya, T Helikar, A Zinovyev, L Calzone, S Cohen-Boulakia, D Thieffry
The CoLoMoTo Interactive Notebook relies on Docker and Jupyter technologies to provide a unified environment to edit, execute, share, and reproduce analyses of qualitative models of biological networks. To date, the framework provides access to software tools including Cell Collective, GINsim, BioLQM, Pint, and MaBoSS. A Python interface has been developed for each of these tools to offer a seamless integration in the Jupyter web interface and ease the chaining of complementary analyses.
Alberto Valdeolivas (I2M) - A Multiplex Network approach to Premature Aging Diseases.
Premature aging (PA) syndromes are a group of heterogeneous rare disorders that recapitulate some of the aspects associated to physiological aging. They are caused by mutations in several genes involved in different biological processes. Genes and proteins do not act isolated in cells but rather interact in complex networks of molecular interactions. In this context, we undertook a network approach to better understand the etiology and pathophysiolgy of these diseases.
First, we extracted the network modules surrounding genes mutated in PA diseases, to define the landscape of biological processes that might be perturbed. To this goal, we applied a strategy based on our recently developed random walk (RW) with restart on multiplex networks . This allows us to navigate and extract information from different layers of physical and functional interactions (e.g., protein-protein, co-expression, molecular complexes) outperforming single-network approaches . We captured modules representing the hallmarks of physiological aging, and compared the processes commonly perturbed in PA diseases, as well as those specific to a subset of diseases.
In a second part, we are developing a strategy to analyse the impact on networks of PA disease-causing mutations. To this goal, we are performing targeted attacks, removing from the multiplex network either genes (to simulate loss-of-function) or some of their interactions (to simulate "edgetic" mutations). A modified version of our RW algorithm allows us to study the topological modifications of the network after the attack, pinpointing to the most affected genes, modules and processes.
1. Valdeolivas,A. et al. Random Walk With Restart On Multiplex And Heterogeneous Biological Networks. 2017. bioRxiv.