Accurate Generation of Conformational Ensembles for Intrinsically Disordered Proteins with IDPFold

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Intrinsically disordered proteins (IDPs) are crucial for biological functions but linked to diseases like cancer and Alzheimer's.
• Studying IDPs is challenging due to their dynamic structures, limited experimental data, and computational costs of traditional methods like molecular dynamics (MD) simulations.
• Current structure prediction methods struggle with IDPs because of poor sequence conservation and scarce experimental characterization.

Purpose of the Study:

• To introduce IDPFold, a novel method for generating conformational ensembles of IDPs directly from their amino acid sequences.
• To overcome the reliance on multiple sequence alignments (MSA) or experimental data for IDP structure prediction.
• To enable a more comprehensive characterization of structural features within IDP ensembles.

Main Methods:

• Utilized fine-tuned diffusion models to predict IDP conformational ensembles.
• Developed IDPFold to directly process protein sequences without requiring MSAs or experimental data.
• Evaluated IDPFold performance across 27 diverse IDP systems.

Main Results:

• IDPFold achieved a Radius of Gyration (Rg) error of -0.06 and an RMSD of 0.65 ppm on Cα secondary chemical shifts against experimental values.
• Demonstrated significantly superior performance compared to existing generative deep learning approaches for IDP structure prediction.
• Successfully generated detailed structural features for IDP ensembles.

Conclusions:

• IDPFold offers a powerful, sequence-based approach for characterizing IDP conformational ensembles.
• The method effectively addresses the limitations of traditional simulation and prediction techniques for IDPs.
• IDPFold facilitates deeper insights into the sequence-disorder-function paradigm, crucial for understanding IDP-related diseases.