Now it is quite usual to use real computers to simulte virtual cells. I suggest that real cells (e.g. cells cultured in vitro) might be considered and used as molecular automata. As an imaginary experience, a molecular automata can be built, using real cells and a chemical inert molecule. I suggest that one could be able to test statistical properties of a 2D gas trapped in a box using this sort of automata. Moreover, I would conjecture that any possible algorithm can be implemented in such molecular automata.