Résumé

Logical modelling constitutes a flexible framework to build qualitative predictive models, which can be readily analysed or simulated as such, and potentially used as scaffolds to build more quantitative (continuous or stochastic) models. We use Multi-valued Decision Diagrams to implement (multi-level) logical updating rules in the modelling software GINsim [1, 2]. This representation enabled the development of efficient algorithms for the identification of stable states, or yet to identify specific (positive or negative) regulatory circuits involved in specific dynamical properties (e.g., multiple attractors or sustained oscillations). To cope with larger molecular networks, we have implemented a flexible reduction method conserving the attractors of the original model into our software GINsim [3, 4]. Furthermore, we have delineated an incremental, compositional strategy to build large models by combining logical models for simpler regulatory modules. In collaboration with Thomas Graf (Center for Genomic Regulation, Barcelona, Spain), we are currently developing a model accounting for the differentiation of common lymphocyte/myoloid progenitors into lymphocyte and macrophage progenitors. This model also recapitulates the reprogramming of T or B lymphocyte precursors into macrophages upon induced expression of the transcription factors CEBPa or CEBPb [5, 6]. Finally, this model can be used to simulate the behaviour of the system in novel situations, e.g. in the presence of combinations of perturbations such as gene knockouts or ectopic expressions.