Inverse engineering of phase separating bio-condensates

Liquid-liquid phase separation (LLPS) can result in complex phase diagrams even when the components comprising a fluid experience simple pairwise interactions. However, these interactions may not be independent of one another, since they arise from a limited set of physicochemical molecular features. Inspired by the observation that intracellular protein/RNA mixtures phase separate into many immiscible membraneless organelles, we seek to design pairwise interactions that can arise from a small number of physicochemical molecular features and yet give rise to complex phase diagrams. We first propose an algorithm for determining the minimum number of physicochemical features required to establish a target phase diagram within a mean-field theory. We then demonstrate the validity of our approach by designing polymer sequences and patchy colloidal particles with a small number of monomer and patch types, respectively, and verifying that the target phase diagram is realized. Our approach provides a principled way to explore the relationship between the physicochemical properties of biomolecules and intracellular LLPS, as well as a systematic design strategy for engineering complex biomolecular condensates de novo.