Unmasking AlphaFold to integrate experiments and predictions in multimeric complexes.

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• AlphaFold has advanced protein structure prediction, particularly for multimeric proteins, as seen in CASP15.
• Limitations remain in predicting large protein complexes and integrating experimental data into prediction pipelines.

Purpose of the Study:

• To introduce AF_unmasked, a novel method to overcome limitations in predicting large protein complexes.
• To enable the integration of experimental data for more confident and accurate structure prediction.

Main Methods:

• Development of AF_unmasked for integrating experimental information into protein structure prediction.
• Application of AF_unmasked for predicting large or difficult protein assemblies and for structural inpainting.

Main Results:

• AF_unmasked successfully builds large protein assemblies with high confidence by integrating experimental data.
• High-quality structures (DockQ score > 0.8) are generated, even with scarce evolutionary information or imperfect experimental inputs.
• The method effectively performs structural inpainting, potentially offering insights into protein dynamics.

Conclusions:

• AF_unmasked provides an efficient and user-friendly approach for predicting large protein complexes with enhanced confidence.
• The method effectively augments and interprets experimental data, advancing structural biology research.