Structure and dynamics of biomolecular condensates revealed by deep-learning-enhanced interferometric microscopy

Biomolecular condensates are submicrometer structures formed through weak, multivalent interactions, primarily involving low complexity domains of intrinsically disordered proteins. Their functions depend sensitively on physical properties such as size, internal protein concentration, and interfacial tension, yet direct measurements of these properties at small scales remain limited. In this work, we use deep learning-enhanced interferometric imaging to quantify size, protein concentration, and interface sharpness in submicrometer condensates formed by the RNA-binding protein DDX4-LCD. Within the two-phase region, DDX4-LCD forms spherical condensates whose internal concentration can be slightly modulated by salt. Additionally, we observe multiple coexisting particle populations, including structures that persist outside the two-phase region and differ in interface properties and dynamics–suggesting a more complex phase diagram than previously assumed.