Biological molecular machines can process information to reduce energy losses.

Biological molecular machines are enzymes that simultaneously catalyze two processes, one donating free energy and second accepting it. Recent studies show that stochastic dynamics of many protein enzymes is characterized by a long-time memory and the possibility of realizing catalyzed processes in various randomly selected ways. Employing a critical complex network, a sample model of such dynamics is specified and investigated using computer simulations. For models of this type, we prove the generalized fluctuation theorem, in which the entropy reduction is possible at the expense of some information creation. Creation and storage of information take place in the transient nonergodic stages of the dynamic process before the completion of free energy transduction cycle. The free energy supply at the beginning of the next cycle erases this information. The transient reduction of entropy reduces energy losses and may prove to be crucial for the movement of proteins on their tracks and the reason for most protein machines to operate as dimers or higher organized assemblies. From a broader physical perspective, the division of free energy into the useful energy and a second function of thermodynamic state, we refer to as organization, is worth emphasizing. Information can be assigned a physical meaning of a change in the value of both these functions.