Protein complexes in cells are modified by enzymes — phosphorylated, methylated, acetylated. Each subunit of a complex can be in one of several states, and enzymes drive transitions between states. The standard view treats these modifications as signals: a kinase phosphorylates a substrate, changing its activity, which triggers a downstream response. The protein complex is a relay in a signaling network.
Kocka, Husain, and Agudo-Canalejo (arXiv:2508.15603, updated March 2026) show that in the strongly driven limit, nonequilibrium protein complexes map exactly onto stochastic, asynchronous cellular automata. Each subunit is a cell in the automaton. Each enzyme set defines a different update rule — a mapping from the states of neighboring subunits to the transition probability of the focal subunit. The enzymatic reactions are the rules; the protein complex is the lattice; the nonequilibrium drive is the clock.
The mapping is not metaphorical. The authors systematically enumerate all possible dynamical rules that arise from different enzyme configurations and identify which rules are capable of implementing finite-state machines. Error-resistant memory appears via multistable attractors — configurations where the complex stays in one of several stable states, robust to thermal fluctuations because the enzymatic drive reinforces the attractor against noise. Prolonged transients function as molecular timekeeping — the complex passes through a predictable sequence of states before settling, and the duration of the transient is a clock.
The reversal: molecular computation does not require specialized information-processing molecules. Generic protein complexes with enzymatic drive already have the computational structure. The complex does not need to be designed as a computer — the computation is a necessary consequence of driven multi-state dynamics on a connected lattice. Any protein complex with three or more subunits, driven by enzymes out of equilibrium, implements some cellular automaton rule. Whether the rule is useful for computation depends on the specific enzymes, but the computational structure is always present.
The structural observation: the computational capacity of a protein complex is not engineered but inherent. The same physics that drives enzymatic modification — nonequilibrium transitions between discrete states on a lattice — is the physics that implements cellular automata. The biology does not need to build a computer. It already is one. The question is which programs it runs, not whether it can compute.