Modeling biomolecular condensates across scales: atomistic, coarse-grained, and data-driven approaches

Abstract

Biomolecular condensates are integral to processes underlying cellular function and dysfunction, and they also present a versatile platform for engineering living cells. Understanding how molecular interactions give rise to condensate form and function is, therefore, a major area of research. Computational modeling has emerged as a powerful tool for uncovering the biophysical principles underlying condensates. Because condensate biophysics spans multiple spatiotemporal scales, from interactions of amino acid side chains to the emergent material properties of entire condensates, decoding their behavior requires multiscale strategies. In this review, we discuss three core classes of computational modeling approaches that extend our ability to probe condensates. We first examine atomistic modeling, which enables a detailed examination of interactions that encode condensate behaviors. We then discuss coarse-grained modeling, with a focus on residue-resolution models, which advance our ability to predict condensate properties with both precision and efficiency. Finally, we summarize advances in data-driven and machine-learning approaches, which leverage molecular simulations to map sequence–property relationships of condensates at a fraction of the cost. Throughout the review, we highlight the key ingredients of each approach, the types of simulations and modeling strategies employed, and the primary observables that can be measured. In doing so, we aim for this review to serve as both an informative and practical guide for leveraging computational approaches to understand and engineer biomolecular condensates.