Loïc Paulevé

Dynamical properties of Discrete Reaction Networks

Reaction networks are commonly used to model the dynamics of populations subject to transformations that follow an imposed stoichiometry. This paper focuses on the efficient characterisation of dynamical properties of Discrete Reaction Networks (DRNs). DRNs can be seen as modeling the underlying discrete nondeterministic transitions of stochastic models of reaction networks. In that sense, a proof of non-reachability in a given DRN has immediate implications for any concrete stochastic model based on that DRN, independent of the choice of kinetic laws and constants. Moreover, if we assume that stochastic kinetic rates are given by the mass-action law (or any other kinetic law that gives non-vanishing probability to each reaction if the required number of interacting substrates is present), then reachability properties are equivalent in the two settings. The analysis of two types of global dynamical properties of DRNs is addressed: irreducibility, i.e., the ability to reach any discrete state from any other state; and recurrence, i.e., the ability to return to any initial state. Our results consider both the verification of such properties when species are present in a large copy number, and in the general case. The necessary and sufficient conditions obtained involve algebraic conditions on the network reactions which in most cases can be verified using linear programming. Finally, the relationship of DRN irreducibility and recurrence with dynamical properties of stochastic and continuous models of reaction networks is discussed.

A generic abstract machine for stochastic process calculi

This paper presents a generic abstract machine for simulating a broad range of process calculi with an arbitrary reaction-based simulation algorithm. The abstract machine is instantiated to a particular calculus by defining two functions: one for transforming a process of the calculus to a set of species, and another for computing the set of possible reactions between species. Unlike existing simulation algorithms for chemical reactions, the abstract machine can simulate process calculi that generate potentially unbounded numbers of species and reactions. This is achieved by means of a just-in-time compiler, which dynamically updates the set of possible reactions and chooses the next reaction in an iterative cycle. As a proof of concept, the generic abstract machine is instantiated for the stochastic pi-calculus, and the instantiation is implemented as part of the SPiM stochastic simulator. The structure of the abstract machine facilitates a significant optimisation by allowing channel restrictions to be stored as species complexes. We also present a novel algorithm for simulating chemical reactions with general distributions, based on the Next Reaction Method of Gibson and Bruck. We use our generic framework to simulate a stochastic pi-calculus model of plasmid co-transfection, where plasmids can form aggregates of arbitrary size and where rates of mRNA degradation are non-exponential. The example illustrates the flexibility of our framework, which allows an appropriate high-level language to be paired with the required simulation algorithm, based on the biological system under consideration.

Under-Approximating cut sets for reachability in large scale automata networks

In the scope of discrete finite-state models of interacting components, we present a novel algorithm for identifying sets of local states of components whose activity is necessary for the reachability of a given local state. If all the local states from such a set are disabled in the model, the concerned reachability is impossible. Those sets are referred to as cut sets and are computed from a particular abstract causality structure, so-called Graph of Local Causality, inspired from previous work and generalised here to finite automata networks. The extracted sets of local states form an under-approximation of the complete minimal cut sets of the dynamics: there may exist smaller or additional cut sets for the given reachability. Applied to qualitative models of biological systems, such cut sets provide potential therapeutic targets that are proven to prevent molecules of interest to become active, up to the correctness of the model. Our new method makes tractable the formal analysis of very large scale networks, as illustrated by the computation of cut sets within a Boolean model of biological pathways interactions gathering more than 9000 components.

Static Analysis of Boolean Networks Based on Interaction Graphs: A Survey

Boolean networks are discrete dynamical systems extensively used to model biological regulatory networks. The dynamical analysis of these networks suffers from the combinatorial explosion of the state space, which grows exponentially with the number n of components. To face this problem, a classical approach consists in deducing from the interaction graph of the network, which only contains n vertices, some information on the dynamics of the network. In this paper, we present results in this topic, mainly by focusing on the influence of positive and negative feedbacks.

Characterization of Reachable Attractors Using Petri Net Unfoldings

Attractors of network dynamics represent the long-term behaviours of the modelled system. Their characterization is therefore crucial for understanding the response and differentiation capabilities of a dynamical system. In the scope of qualitative models of interaction networks, the computation of attractors reachable from a given state of the network faces combinatorial issues due to the state space explosion.

Boolean network identification from perturbation time series data combining dynamics abstraction and logic programming

Boolean networks (and more general logic models) are useful frameworks to study signal transduction across multiple pathways. Logic models can be learned from a prior knowledge network structure and multiplex phosphoproteomics data. However, most efficient and scalable training methods focus on the comparison of two time-points and assume that the system has reached an early steady state. In this paper, we generalize such a learning procedure to take into account the time series traces of phosphoproteomics data in order to discriminate Boolean networks according to their transient dynamics. To that end, we identify a necessary condition that must be satisfied by the dynamics of a Boolean network to be consistent with a discretized time series trace. Based on this condition, we use Answer Set Programming to compute an over-approximation of the set of Boolean networks which fit best with experimental data and provide the corresponding encodings. Combined with model-checking approaches, we end up with a global learning algorithm. Our approach is able to learn logic models with a true positive rate higher than 78% in two case studies of mammalian signaling networks; for a larger case study, our method provides optimal answers after 7min of computation. We quantified the gain in our method predictions precision compared to learning approaches based on static data. Finally, as an application, our method proposes erroneous time-points in the time series data with respect to the optimal learned logic models.

Cell cycle progression is regulated by intertwined redox oscillators

The different phases of the eukaryotic cell cycle are exceptionally well-preserved phenomena. DNA decompaction, RNA and protein synthesis (in late G1 phase) followed by DNA replication (in S phase) and lipid synthesis (in G2 phase) occur after resting cells (in G0) are committed to proliferate. The G1 phase of the cell cycle is characterized by an increase in the glycolytic metabolism, sustained by high NAD+/NADH ratio. A transient cytosolic acidification occurs, probably due to lactic acid synthesis or ATP hydrolysis, followed by cytosolic alkalinization. A hyperpolarized transmembrane potential is also observed, as result of sodium/potassium pump (NaK-ATPase) activity. During progression of the cell cycle, the Pentose Phosphate Pathway (PPP) is activated by increased NADP+/NADPH ratio, converting glucose 6-phosphate to nucleotide precursors. Then, nucleic acid synthesis and DNA replication occur in S phase. Along with S phase, unpublished results show a cytosolic acidification, probably the result of glutaminolysis occurring during this phase. In G2 phase there is a decrease in NADPH concentration (used for membrane lipid synthesis) and a cytoplasmic alkalinization occurs. Mitochondria hyperfusion matches the cytosolic acidification at late G1/S transition and then triggers ATP synthesis by oxidative phosphorylation. We hypothesize here that the cytosolic pH may coordinate mitochondrial activity and thus the different redox cycles, which in turn control the cell metabolism.

Stochastic simulation of multiple process calculi for biology

Numerous programming languages based on process calculi have been developed for biological modelling, many of which can generate potentially unbounded numbers of molecular species and reactions. As a result, such languages cannot rely on standard reaction-based simulation methods, and are generally implemented using custom stochastic simulation algorithms. As an alternative, this paper proposes a generic abstract machine that can be instantiated to simulate a range of process calculi using a range of simulation methods. The abstract machine functions as a just-in-time compiler, which dynamically updates the set of possible reactions and chooses the next reaction in an iterative cycle. We instantiate the generic abstract machine with two Markovian simulation methods and provide encodings for four process calculi: the agent-based pi-calculus, the compartment-based bioambient calculus, the rule-based kappa calculus and the domain-specific DNA strand displacement calculus. We present a generic method for proving that the encoding of an arbitrary process calculus into the abstract machine is correct, and we use this method to prove the correctness of all four calculus encodings. Finally, we demonstrate how the generic abstract machine can be used to simulate heterogeneous models in which discrete communicating sub-models are written using different domain-specific languages and then simulated together. Our approach forms the basis of a multi-language environment for the simulation of heterogeneous biological models.

Under-approximation of Reachability in Multivalued Asynchronous Networks

The Process Hitting is a recently introduced framework designed for the modelling of concurrent systems. Its originality lies in a compact representation of both components of the model and its corresponding actions: each action can modify the status of a component, and is conditioned by the status of at most one other component. This allowed to define very efficient static analysis based on local causality to compute reachability properties. However, in the case of cooperations between components (for example, when two components are supposed to interact with a third one only when they are in a given configuration), the approach leads to an over-approximated interleaving between actions, because of the pure asynchronous semantics of the model. To address this issue, we propose an extended definition of the framework, including priority classes for actions. In this paper, we focus on a restriction of the Process Hitting with two classes of priorities and a specific behaviour of the components, that is sufficient to tackle the aforementioned problem of cooperations. We show that this class of Process Hitting models allows to represent any Asynchronous Discrete Networks, either Boolean or multivalued. Then we develop a new refinement for the under-approximation of the static analysis to give accurate results for this class of Process Hitting models. Our method thus allows to efficiently under-approximate reachability properties in Asynchronous Discrete Networks; it is in particular conclusive on reachability properties in a 94 components Boolean network, which is unprecedented.

Goal-Oriented Reduction of Automata Networks

We consider networks of finite-state machines having local transitions conditioned by the current state of other automata. In this paper, we introduce a reduction procedure tailored for reachability properties of the form “from global state \({s}\), there exists a sequence of transitions leading to a state where an automaton g is in a local state \(\top \)”. By analysing the causality of transitions within the individual automata, the reduction identifies local transitions which can be removed while preserving all the minimal traces satisfying the reachability property. The complexity of the procedure is polynomial with the total number of local transitions, and exponential with the maximal number of local states within an automaton. Applied to Boolean and multi-valued networks modelling dynamics of biological systems, the reduction can shrink down significantly the reachable state space, enhancing the tractability of the model-checking of large networks.