The dynamic ensembles of intrinsically disordered regions (IDRs) are fundamental to their function yet designing them with specific behaviors remains a major challenge. Current AI models and coarse-grained simulations fail to reliably predict the specific intramolecular interactions that define an IDR’s transient structure. To address this gap, we developed the Interaction Map, a framework that decodes an IDR’s ensemble by normalizing all-atom simulation data against an ideal polymer model. This method distills complex contact data into an intuitive visualization of key attractive and repulsive long-range interactions. We demonstrate its power by linking function-altering mutations in p53 to specific changes in its interaction landscape and by accurately predicting the solution sensitivity of IDR ensembles. Leveraging this insight, we developed a “mutation scanning” method for rational protein design, successfully engineering IDR mutants of UPF2 and TDG with altered and predictable solution sensitivities. To make this tool accessible, we also deployed a version based on coarse-grained simulations on Google Colab for experimentalists. The Interaction Map is not only a powerful tool for engineering IDR function but also serves as a critical encoding method to guide the next generation of AI models for predicting dynamic protein ensembles.
Yu et al. (Sun,) studied this question.