AlphaFold-Multimer accurately captures interactions and dynamics of intrinsically disordered protein regions

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Intrinsically disordered protein regions (IDRs) present significant challenges for structural characterization due to their dynamic and versatile nature.
• Recent advancements in protein structure prediction offer new opportunities to study IDR interactions.

Purpose of the Study:

• To evaluate the capability of AlphaFold-Multimer in accurately predicting interactions involving intrinsically disordered protein regions (IDRs).
• To assess the reliability of AlphaFold-Multimer's predictions for various IDR binding modes and dynamics.

Main Methods:

• Utilized multiple datasets encompassing diverse IDR binding modes to probe AlphaFold-Multimer's performance.
• Analyzed AlphaFold-Multimer's intrinsic scores (e.g., Predicted Aligned Error, residue-ipTM) to assess prediction accuracy and distinguish between true and decoy interactions.
• Benchmarked predictions using full-length proteins versus cognate IDRs.

Main Results:

• AlphaFold-Multimer successfully predicts various bound IDR structures with high accuracy.
• Intrinsic scores effectively differentiate reliable predictions from decoys and assess structural heterogeneity.
• Prediction quality decreases for fuzzy interactions, but specific scores correlate with structural heterogeneity, enabling distinction between binding modes.
• Predictions using full-length proteins are less accurate than those using cognate IDRs.

Conclusions:

• AlphaFold-Multimer is a powerful tool for predicting IDR interactions and their engagement modes.
• The study introduces "minD" to identify potential IDR interaction sites in full-length proteins.
• Reliable prediction of IDR interactions is achievable with careful use of AlphaFold-Multimer's scoring metrics.