Research projects – Team O. Radulescu
RESEARCH PROJECTS
We are developing mathematical approaches for learning and analysing mechanistic models of biological systems at various levels of organisation, with a focus on the host-pathogen interactions in several infectious diseases (malaria, HIV) and on cancer.
Mathematical modelling of biological systems, in their full details, is a daunting challenge. In order to cope realistically with the dynamics of molecular pathways and gene networks in the cell, bottom-up models use thousands of variables. Furthermore, in models of tissues, populations of cells with complex single cell dynamics must be described collectively within a spatially heterogeneous framework. In order to cope with this complexity, we develop rigorous and automated methods for generating hierarchies of simplified models that keep, at each scale, only essential processes and components. Our modelling approaches provide solutions to many problems in fundamental biology and medicine.
We also develop novel AI methods for extracting information from biological data. Our vision in this field is to combine data driven “black box” models with knowledge driven “white box” models within hybrid AI approaches.
Theme 1
LARGE REGULATORY NETWORKS: FROM MOLECULAR INTERACTIONS TO BIOLOGICAL FUNCTION
We develop mathematical methods for reconstruction and analysis of large biochemical networks involved in cellular signalling and metabolism. These networks are described as complex systems of interacting molecules, together with their dynamics in space and in time. Although our approaches can be applied to the understanding of regulatory processes in all organisms, we study with particular emphasis networks of higher eukaryotes, involved in systems biology of human health and disease. To improve the effectiveness of mathematical modelling we take into account and benefit from notable properties of biological regulation networks such as modularity, multiscaleness, and robustness. Some recent developments and projects: i) cross-talk of signaling pathways in cancer (Ewing sarcoma, cervical cancer); ii) cell cycle hybrid modeling; iii) lipid metabolism in various species (fatty acid balance in mice liver during fasting, phospholipid synthesis in Plasmodium falciparum); iv) canalization of early development stages in dipteran insects; v) robustness of complex regulatory networks by dimension compression; vi) stochastic networks.
Theme 2
MULTISCALE APPROACHES: FROM INDIVIDUAL MOLECULE AND MOLECULAR INTERACTIONS TO VIRTUAL CELL
This goal combines 1) and 2). By using across scales descriptions, we aim to integrate in our models both physical processes and regulatory networks. Our methodology can be described as hierarchical modelling. It allows descriptions of biological systems at different scales and it is based on model reduction and model conversion techniques. By setting solid theoretical basis for hierarchical modelling, we contribute to the larger international effort endeavoring the future creation of an integrated model of the whole cell (virtual cell).
We are developing mathematical approaches for learning and analysing mechanistic models of biological systems at various levels of organisation, with a focus on the host-pathogen interactions in several infectious diseases (malaria, HIV) and on cancer.
Mathematical modelling of biological systems, in their full details, is a daunting challenge. In order to cope realistically with the dynamics of molecular pathways and gene networks in the cell, bottom-up models use thousands of variables. Furthermore, in models of tissues, populations of cells with complex single cell dynamics must be described collectively within a spatially heterogeneous framework. In order to cope with this complexity, we develop rigorous and automated methods for generating hierarchies of simplified models that keep, at each scale, only essential processes and components. Our modelling approaches provide solutions to many problems in fundamental biology and medicine.
We also develop novel AI methods for extracting information from biological data. Our vision in this field is to combine data driven “black box” models with knowledge driven “white box” models within hybrid AI approaches.
Some recent developments and projects :
- Mechanical instabilities of tubular lipid membranes, protein/protein and protein/membrane interactions,
- collective dynamical properties of low-dimensional lattices gas mimicking motor protein intracellular processes;
- tubuline and microtubule response under a hydrostatic pressure;
- nucleation growth and maturation of Mycobacterium smegmatis biofilms,
- in silico description of fluorescent microscopy and spectroscopy experiments.
Past projects :
- non-equilibrium and stochastic properties of a single molecular motor;
- mechanisms of force production and transport of acto-myosin systems under ATP depletion.
A few projects:
Formal methods in systems biology (project SYMBIONT) https://systems-biology-lphi.cnrs.fr/formal/
Stochastic dynamics of gene expression https://systems-biology-lphi.cnrs.fr/stochastic/
Machine learning and AI methods for systems biology https://systems-biology-lphi.cnrs.fr/AI/
Cancer heterogeneity and resistance to treatment https://systems-biology-lphi.cnrs.fr/cancer/
Molecular aspects of infection https://systems-biology-lphi.cnrs.fr/infection/