Predicting biomolecular phase separation with field-theoretic simulations

The phase separation of biomolecules has catalyzed a surge of computational models that seek to predict phase separation from monomer sequence alone. Many recent studies have relied on particle-based models, yet computational limitations have restricted these models to approximations such as implicit solvents, Debye-Hückel electrostatics and simplistic treatments of phase equilibrium. In this work, we present a new model for biomolecular phase separation that avoids these approximations by using field-theoretic simulations. We demonstrate that field-theoretic simulations can be constructed to be formally equivalent to particle-based simulations and that both types of simulations yield identical values for the pressure and the chemical potential. Next we compare the performance of particle vs field simulations and show that field-theoretic simulations converge several orders of magnitude more quickly, despite giving identical results. This significant speedup permits our model to efficiently perform detailed biomolecular simulations with explicit solvent, detailed electrostatics, and thermodynamically rigorous phase diagrams using the Gibbs ensemble. Our model can recapitulate recent experimental data on intrinsically-disordered proteins and can examine the effects of amino acid sequence on their phase separation. Taken together, this work demonstrates that field-theoretic simulations can unlock a detailed molecular view into the physics of biomolecular phase separation.