Discriminating Molecular Mixtures with Phase Separation

The cytoplasm, nucleus, & plasma membrane contain order thousands of distinct biomolecules in concentrations ranging from single-copy to millimolar. These mixtures are often prone to de-mixing transitions where components phase-separate into co-existing liquid phases. Individual components vary across molecular features that influence phase behavior like charge, domain identity, and phosphorylation state. Resolving these molecular features from macroscopic phase behavior is appealing, but the feasibility and reliability of such schemes has not been tested. Within a mean-field theory and lattice Monte-Carlo simulations we quantify the ability of mechanisms from simple binding to phase-separation to discriminate between increasingly subtle molecular distributions. Phase-separation is naturally sensitive to the entire molecular feature distribution while simple mass action models can only report low-order information. Incorporating additional physics, like finite valency constraints, amplifies performance on certain tasks but degrades performance on others. To solve more complex discriminatory problems we exploit the kinetics of phase-separation. These 'kinetic solutions' take advantage of the entire demixing pathway and are often more compact than their equilibrium counterparts.