Information and Thermodynamics in Biochemical Systems

Thomas Ouldridge (Imperial)

Living cells use readout molecules to record the state of receptors that detect ligands in their environment. This process appears to be similar to measurements made by computational devices, as extensively studied in the literature following Maxwell's demon. But at what level do measurements made by cellular systems map onto computational measurements made, for example, by magnetic devices? Can cells reach the thermodynamic limit of minimal dissipation for a given measurement accuracy, and, if not, what is the cause?

We show that, as in canonical descriptions of computation, biochemical networks can induce correlations between non-interacting degrees of freedom. If these correlations are not used to extract work, then they set a lower bound on entropy generation. For general input data, autonomous biochemical networks cannot reach this lower bound, even with arbitrarily slow reactions or under arbitrarily small thermodynamic driving. They face a trade-off between accuracy and dissipation that is qualitatively distinct from and more severe than that required thermodynamically. Nonetheless, the costs can remain close to the thermodynamic bound unless accuracy is extremely high. Furthermore, biomolecular reactions can achieve thermodynamically optimal measurement if the concentration of chemical fuel molecules are manipulated exogenously.

Finally, the fact that generating correlations between non-interacting degrees of freedom is costly suggests that this mutual information can be used as a source of work. We propose a biochemical device to explore the inter-conversion of information and work, and investigate its operation.