Many proteins do more than adopt individual structures: in solution, they interact collectively, remaining dispersed, forming finite clusters, or separating into condensed phases. Predicting these outcomes from molecular sequence and architecture—and determining how the resulting phases nucleate and evolve—is a central challenge in molecular thermodynamics.
In this talk, I will describe our efforts to build a computational framework connecting protein sequence to collective assembly. I will begin with human transcription factors, whose compositional diversity, modular organization, and abundance of intrinsically disordered regions make them a rich model family for mapping sequence space. I will show how interpretable sequence- and domain-level representations reveal organization across the human transcription-factor landscape, and how selected sequence features translate into differences in phase behavior and assembly propensity.
I will then turn to the formation of condensed phases from dilute solution, using nucleation free-energy landscapes to identify the barriers that govern their emergence. Finally, I will show how continuum descriptions can extend this molecular picture to condensate evolution and coarsening over timescales beyond direct simulation. Together, these approaches connect sequence, molecular interactions, phase behavior, and dynamics toward a predictive framework for understanding—and ultimately controlling—protein condensation.