Percolation in protein cores: a novel approach to protein decoy detection

Protein cores are regions of densely packed, solvent-excluded residues, and void space inside of cores can often destabilize the structure. Here, we measure the void space in protein cores and find that the void structure is equivalent to that in jammed packings of repulsive residue-shaped particles. A continuum void percolation transition can be defined as when the characteristic void length scale approaches the system size. Using finite-size scaling, we show that the percolation of void space in protein cores belongs to same universality class as voids in static packings of residue-shaped particles. This result provides a novel approach for evaluating whether computationally designed protein structures will take a desired fold in experiments. Molecular dynamics simulations often generate “decoy” protein structures that have low potential energy, but are not observed experimentally. We argue that decoy protein structures will have a fundamentally different void distribution, i.e. belong to a different universality class than the void distribution in real protein structures. Therefore, by analyzing the universal aspects of connected voids in computationally generated protein structures, we will be able to differentiate experimentally observed structures from protein decoys.