This thesis describes a method to predict the tertiary structure of RNA molecules from their sequences. It relies on the observation that their folding is hierarchical and modular; it consists, first, in the extraction of those modules (the helices and junctions between helices) and their classification in topological families, then, in, an optimisation step to combine all those autonomous modules into a folded and stable molecule. This folding relies on an algorithmic approach to game theory. We present a modelisation of the folding process as a game, a cost function associated to that game, and several heuristics to search for Nash equilibrium.
