Using AlphaFold2 to Predict the Conformations of Side Chains in Folded Proteins

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• AlphaFold revolutionized protein structure prediction, but its accuracy for individual amino acid side chain conformations is still under investigation.
• Accurate side chain prediction is crucial for molecular modeling, especially for understanding mutation effects on protein stability and ligand binding.

Purpose of the Study:

• To evaluate the ability of ColabFold (an AlphaFold2 implementation) to predict residue side chain conformations in folded proteins.
• To explore the integration of ColabFold with a Potts model for analyzing cooperative mutations and their structural consequences.

Main Methods:

• Assessed ColabFold's side chain prediction accuracy on 10 benchmark proteins, analyzing errors in dihedral angles.
• Investigated the influence of side chain polarity and structural templates on prediction accuracy.
• Employed a Potts sequence-based statistical energy model for large-scale mutational scans and integrated it with ColabFold to predict structural changes.

Main Results:

• ColabFold exhibited average prediction errors of $$ for $$ and $$ for $$ dihedral angles.
• Prediction errors were lower for non-polar side chains and slightly improved with structural templates.
• ColabFold showed a bias towards common rotamer states found in the Protein Data Bank (PDB).

Conclusions:

• ColabFold demonstrates utility in predicting side chain conformations, though with limitations for rare states.
• The combined Potts model and ColabFold pipeline offers a novel approach to investigate the structural impact of cooperative mutations.
• This integrated tool can advance the understanding of the relationship between protein mutations, structural cooperativity, and biological fitness.