Main research projectsThese are the two projects in which I dedicate most of my research time and resources.
Real-time observation and control of biomolecular processes at the single-cell level
Because of significant cell-to-cell variability, classical (population) models show poor prediction capabilities when compared with single cell data. In this project we investigate how much our prediction capabilities can be improved by exploiting single-cell observations in real-time. To do this, we have developed an integrated experimental platform combining microscopy, microfluidics and software. An effective experimental way of testing model quality is to perform model-based control. As a first case study, we consider the well-known HOG signal transduction pathway, involved the hyperosmotic stress response in yeast, and use it to control gene expression in real-time.
This work is done in collaboration with Pascal Hersen ( MSC, CNRS/Paris7) and Jannis Uhlendorf ( INRIA/MSC). Preliminary results have been published in PSB'11 and IFAC'11 conferences. This work is also at the core of a larger projet, Iceberg (ANR Investissement d'Avenir), involving 6 labs: Contraintes (G. Batt), MSC (P. Hersen), BM2A (O. Gandrillon), LIFL (C. Lhoussaine), IJM (R. Veitia), and PPS (J. Krivine).
In this project, we develop methods to support the rational design of artificial tissues. The development of such multicellular systems is a central challenge in synthetic biology. On the design level, the problem is to engineer intracellular processes such that the desired tissue behavior emerges from cell interactions. The analysis -and even more the optimizations- of such multiscale systems calls for the development of abstractions techniques to make the problem computationally tractable.
These methods are applied to the development of a mammalian artificial tissue homeostasis system. The goal is to design in silico and construct in vivo a tissue that autonomously maintins its cell density at a desired level. It requires the development and integration of cell-cell communication and growth control mechanisms.
This work is done in collaboration with Francois Bertaux ( INRIA Bang), Dirk Drasdo ( INRIA Bang), Xavier Duportet ( INRIA Contraintes/MIT), Szymon Stoma ( INRIA Contraintes), and Ron Weiss ( MIT). This work is also part of a larger project, Syne2Arti (ANR Cosinus), additionally involving Alexandre Donze ( Verimag) and Oded Maler ( Verimag).
Secondary projectsThese are ongoing projects coordinated by other members of the Contraintes group or by other close colaborators.
Quantification of biological system robustness
Robustness is the capacity of a system to maintain a function in the face of perturbations. It is essential for the correct functioning of natural and engineered biological systems. Robustness is generally defined in an ad hoc, problem-dependent manner, thus hampering the fruitful development of a theory of biological robustness, recently advocated by Kitano. In this work, we propose a general definition of robustness that applies to any biological function expressible in a language for behavior specification (LTL temporal logic), and to broad model classes and perturbation types. The applicability of our approach is illustrated by testing and improving in silico the robustness of the timed behavior of a synthetic transcriptional cascade.
This work is done in collaboration with Aurélien Rizk, Francois Fages, and Sylvain Soliman (INRIA Contraintes). Results have notably been published in Bioinformatics 2009.
Efficient parameter search methods for hybrid models of gene networks: PMA and PA models
There is currently no efficient methods for the analysis of uncertain non-linear dynamical systems such as those encountered in systems biology. In these works, we focus on two interesting classes of non-linear systems -piecewise affine (PA) and piecewise multiaffine (PMA) systems- and propose automated methods for their analysis. Both model types are globally complex (due to their piecewise nature) but locally simple (affine or multiaffine). This feature makes them particularly well adapted to perform discrete or timed abstractions and then model checking.
For PA models, I work mostly in collaboration with Hidde de Jong and members of the Ibis INRIA group (see e.g. our Bioinformatics 2010 paper). For PMA models, I work mostly with Calin Belta (Boston Univ.) and Ron Weiss (MIT). Recently, the PMA framework was also used for cardiac cell modeling (see the CAV'11 paper with Radu Grosu and colleagues).