A mixture of RNA and protein molecules in solution can form a gel — a system-spanning network held together by specific binding interactions between the two species. Chen, Vishnu, Besenius, Konig, and Schmid simulate this process using coarse-grained molecular dynamics and find that the sol-gel transition is reentrant: increasing the concentration of one component first drives gelation and then, past a second threshold, reverses it.
The mechanism is saturation. Each RNA molecule has a fixed number of binding sites for protein, and each protein has a fixed number for RNA. At low concentrations, molecules find partners and begin assembling into clusters. At intermediate concentrations, the clusters span the simulation box — percolation occurs, and the system gels. But at high concentrations, the binding sites become saturated. Every site is occupied, and molecules can no longer serve as bridges connecting distant parts of the network. The network fragments. The gel dissolves back into a sol.
The Semenov-Rubinstein theory — a mean-field framework developed for associative polymers — predicts the equilibrium binding numbers accurately in the semidilute regime. But it significantly overestimates the concentration range over which percolation is observed. The discrepancy arises because the theory treats the network as a tree — a structure without cycles. Real networks form loops, and loops consume binding sites without extending the network's reach. The mean-field picture misses the cost of redundancy.
The through-claim applies wherever networks form through specific binding. Adding more connectors does not always strengthen the network. Past a critical density, connectors bind to each other instead of bridging distant regions, and the structure that connectivity built begins to dissolve. More connections can mean less coherence — an architectural paradox that mean-field theories, by ignoring loops, are constitutionally unable to see.