Protein complex prediction using Rosetta, AlphaFold, and mass spectrometry covalent labeling

Area of Science:

• Structural biology
• Computational chemistry
• Biophysics

Background:

• Covalent labeling (CL) coupled with mass spectrometry is an analytical technique for studying protein-protein complex structures.
• Existing CL data is often sparse, limiting its ability to unambiguously determine protein structures.
• Computational algorithms are necessary to interpret CL data for structural elucidation.

Purpose of the Study:

• To develop and validate a hybrid computational method for determining protein-protein complex structures using covalent labeling data.
• To assess the accuracy and effectiveness of the integrated approach in distinguishing native-like models.

Main Methods:

• A hybrid method was developed, integrating AlphaFold-generated protein complex subunit models with differential covalent labeling data.
• The method employed CL-guided protein-protein docking within the Rosetta computational framework.
• A benchmark set of protein complexes was used to evaluate the performance of the developed method.

Main Results:

• The integrated approach, utilizing CL data, achieved a root-mean-square deviation (RMSD) below 3.6 Å for 5 out of 5 benchmark complexes.
• In contrast, models generated without CL data only achieved this level of quality for 1 out of 5 complexes.
• The results demonstrate a significant improvement in structural model accuracy when incorporating CL data.

Conclusions:

• The developed hybrid method successfully leverages covalent labeling experimental data to enhance the accuracy of protein-protein complex structure determination.
• This integrated approach effectively distinguishes between native-like and non-native-like structural models.
• The findings highlight the utility of combining computational modeling with experimental data for advancing structural biology.