When B₂ is Not Enough: Unraveling the Limitations of Simple Metrics for Assessing the Phase Behavior of Intrinsically Disordered Proteins

The phase behavior of intrinsically disordered proteins (IDPs) crucially underlies myriad biological processes. Consequently, methods that facilitate understanding and predictions of how protein sequence influences phase behavior are of significant interest. Here, we evaluate the efficacy of several inexpensive metrics to quantify the propensity of IDPs to undergo phase separation using coarse-grained simulation and data-driven analysis of more than 2,000 IDP sequences. We find that a novel metric, which we refer to as "expenditure density," outperforms more traditional assessments (e.g., homotypic second virial coefficient and radius of gyration), offering low computational cost, high accuracy, and quantitative generalizability. This work establishes a foundation for developing more effective machine learning classifiers for IDP phase behavior and further elucidating key sequence determinants.