Accurate Generation of Conformational Ensembles for Intrinsically Disordered Proteins with IDPFold

Intrinsically disordered proteins (IDPs) play pivotal roles in various biological functions whose dynamic structures are closely associated with many human diseases, including cancer, diabetes, and Alzheimer disease. Structural investigations of IDPs typically involve a combination of molecular dynamics (MD) simulations and experimental data to mitigate intrinsic biases in simulation methods. However, the high computational cost of these simulations and the limited availability of experimental data significantly restrict their applicability. Despite the recent advancements in structure prediction for structured proteins, understanding the conformational properties of IDPs remains challenging, partly due to the poor conservation of disordered protein sequences and the scarcity of experimental characterization. Here, IDPFold is introduced as a method capable of generating conformational ensembles for IDPs directly from their sequences using fine-tuned diffusion models. IDPFold eliminates the reliance on multiple sequence alignments (MSA) or experimental data, offering a more detailed characterization of structural features in IDP ensembles. Evaluated across 27 IDP systems, IDPFold achieves Rg error of -0.06 and an RMSD of 0.65 ppm on Cα secondary chemical shifts with experimental values, significantly better than all existing generative deep learning approaches. IDPFold can be used to elucidate the sequence-disorder-function paradigm of IDPs.