Prediction of Protein Complex Structure Using Surface-Induced Dissociation and Cryo-Electron Microscopy

Area of Science:

• Structural biology
• Computational chemistry
• Biophysics

Background:

• Mass spectrometry (MS) techniques, including surface-induced dissociation (SID), provide structural insights into proteins and protein complexes.
• Appearance energies (AE) derived from SID experiments correlate with protein-protein interface structures but are insufficient alone for complex determination.
• Previous computational docking methods required extensive prior structural knowledge.

Purpose of the Study:

• To improve the accuracy of protein complex structure prediction by integrating SID data with computational modeling.
• To develop a robust method using less prior structural information, such as homology models and unbound crystal structures.
• To assess the performance of combined computational and experimental data for predicting protein complex structures.

Main Methods:

• Utilized flexible ensemble docking with input structures requiring less prior knowledge (homology models, unbound/perturbed crystal structures).
• Integrated data from Rosetta, cryo-electron microscopy (cryo-EM), and SID appearance energies (AE) for scoring.
• Employed symmetric docking to construct complete protein complexes using symmetry information.

Main Results:

• Flexible ensemble docking achieved <4 Å RMSD100 for all 15/15 predicted structures when using combined Rosetta, cryo-EM, and SID data, compared to 7/15 without.
• Symmetric docking yielded <4 Å RMSD100 for 14/15 cases with experimental data, significantly outperforming the 5/15 accuracy without SID and cryo-EM.
• A developed confidence metric accurately predicted all 26/26 high-confidence proteins.

Conclusions:

• Combining SID appearance energies, cryo-electron microscopy, and computational docking significantly enhances the accuracy of protein complex structure prediction.
• The developed method effectively utilizes diverse structural inputs, including those with limited prior knowledge, for robust structural modeling.
• The high accuracy of the confidence metric demonstrates the reliability of the integrated approach for structural biology applications.