Evaluation of AlphaFold 3's Protein-Protein Complexes for Predicting Binding Free Energy Changes upon Mutation

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• AlphaFold 3 (AF3) advances protein structure prediction by modeling protein-protein complexes.
• Independent validation is crucial for assessing AF3's utility in drug discovery and protein engineering.

Purpose of the Study:

• To evaluate the accuracy of AF3-predicted protein complex structures.
• To assess AF3's performance in predicting binding free energy changes upon mutation.

Main Methods:

• Utilized the SKEMPI 2.0 database comprising 317 protein-protein complexes and 8338 mutations.
• Applied AF3 complex structures to the MT-TopLap (MultiTask-Topological Laplacian) model.
• Calculated Pearson correlation coefficient and Root Mean Square Error (RMSE) for binding free energy predictions.

Main Results:

• Achieved a Pearson correlation coefficient of 0.86 for predicting binding free energy changes, slightly lower than PDB structures (0.88).
• Observed an 8.6% increase in prediction RMSE compared to PDB complex structures.
• Identified large errors in some AF3 complex structures not reflected by the ipTM metric, and unreliability in intrinsically flexible regions.

Conclusions:

• AF3 complex structures show promise but require careful validation, especially for flexible regions.
• The current AF3 predictions are not fully reliable for applications involving intrinsically disordered proteins or flexible domains.
• Further refinement of AF3 is needed to improve accuracy and reliability for comprehensive protein structure prediction.