Automated navigation of condensate phase behavior with active machine learning

Abstract

Abstract Biomolecular condensates are essential cellular structures formed via biomacromolecule phase separation. Synthetic condensates allow for systematic engineering and understanding of condensate formation mechanisms and to serve as cell-mimetic platforms. Phase diagrams give comprehensive insight into phase separation behavior, but their mapping is time-consuming and labor-intensive. Here, we present an automated platform for efficiently mapping multi-dimensional condensate phase diagrams. The automated platform incorporates a pipetting system for sample formulation and an autonomous confocal microscope for particle property analysis. Active machine learning is used for iterative model improvement by learning from previous results and steering subsequent experiments towards efficient exploration of the binodal. The versatility of the pipeline is demonstrated by showcasing its ability to rapidly explore the phase behavior of various polypeptides, producing detailed and reproducible multidimensional phase diagrams. The self-driven platform also quantifies key condensate properties such as particle size, count, and volume fraction, adding functional insights to phase diagrams.