Benchmarking AlphaFold-Generated Structures of Chemokine-Chemokine Receptor Complexes

Area of Science:

• Structural Biology
• Computational Biology
• Biochemistry

Background:

• AlphaFold and AlphaFold-Multimer are leading tools for protein structure prediction.
• Accurate modeling of protein-protein interactions, like chemokine-receptor binding, is crucial for understanding biological processes.
• Experimental determination of these structures can be challenging.

Purpose of the Study:

• To benchmark AlphaFold-Multimer's performance in predicting chemokine-chemokine receptor structures.
• To develop a customizable prediction pipeline (LIT-AlphaFold) for enhanced input parameter control.
• To investigate the impact of template and multiple sequence alignment parameters on prediction accuracy.

Main Methods:

• Extensive benchmarking of AlphaFold-Multimer against experimentally determined chemokine-chemokine receptor structures.
• Development and utilization of the LIT-AlphaFold pipeline for customized structure prediction.
• Analysis of global model quality and specific structural features relevant to chemokine recognition.

Main Results:

• AlphaFold-Multimer accurately predicted differences in chemokine binding orientations.
• The unique binding of the CXCL12-ACKR3 complex was correctly reproduced.
• Predictions of the N-terminus offered insights into a potential chemokine recognition site.
• A high-confidence model for the CXCL12-CXCR4 complex was generated, aligning with experimental data.

Conclusions:

• AlphaFold-Multimer is a valuable tool for modeling chemokine-chemokine receptor complexes.
• LIT-AlphaFold provides enhanced control and customization for structure prediction.
• The study successfully modeled previously unsolved complexes and provided insights into binding mechanisms.