Section: New Results
The chemical reaction metaphor describes computation in terms of a chemical solution in which molecules (representing data) interact freely according to reaction rules (representing the program). Chemical programs can be formalized as associative-commutative rewritings (reactions) of multisets (chemical solutions). This model of computation is well-suited to the specification of complex computing infrastructures. In particular, the orderless interactions between elements that occur in large parallel or open systems as well as autonomicity (e.g. self-healing, self-protection, self-optimization, etc.) are naturally expressed as reaction rules. We have described classical coordination mechanisms and parallel programming models (Linda, Petri Nets, Kahn Networks) in the same chemical setting  . All these examples put forward the simplicity and expressivity of the chemical paradigm.
A drawback of chemical languages is that their very high-level nature usually leads to very inefficient programs. We are currently looking at approaches to refine chemical programs to more efficient lower-level programs. The idea is to specify separately the data structures, the selection of elements and the scheduling of rules using domain specific languages. The goal is to use these additional components to automatically refine chemical programs into C-like programs. The overall approach is related to aspect-oriented programming where the chemical program represents the base functionality and the other components can be seen as implementation aspects.
This line of research is followed by Marnes Hoff in his PhD thesis. It takes place within the AutoChem project (see Section 8.2.1 ).
Component-based modeling and reachability analysis of genetic networks
Participant : Gregor Gössler.
Genetic regulatory networks usually encompass a multitude of complex, interacting feedback loops. Being able to model and analyze their behavior is crucial for understanding the interactions between the proteins, and their functions. Genetic regulatory networks have been modeled as discrete transition systems by many approaches, benefiting from a large number of formal verification algorithms available for the analysis of discrete transition systems. However, most of these approaches face the problem of state space explosion, as even models of modest size (from a biological point of view) usually lead to large transition systems, due to a combinatorial blow-up of the number of states. This problem has been addressed with the component-based approach of  — based on the mathematically well-founded formalism of qualitative simulation  — where the discrete abstraction is constructed and analyzed modularly, allowing to deal with complex, high-dimensional systems.
We have further improved this technique by allowing for a more precise, conservative abstraction, and provided both correctness and completeness results  .
We are currently working, in cooperation with H. de Jong (Ibis team from Grenoble) and G. Batt (Contraintes team from Rocquencourt), on parametric models of genetic networks that reflect the lack of knowledge about the position of focal points with respect to the thresholds, and the ordering of thresholds. The goal is to determine automatically feasible parameter values corresponding to an observed or desired behavior.