Protein oligomer structure prediction using GALAXY in CASP14.

Area of Science:

• Structural biology
• Computational biology
• Bioinformatics

Background:

• Protein interactions and oligomerization are fundamental to biological functions.
• Accurate prediction of protein oligomer structures is essential for understanding protein roles.
• Computational methods offer a valuable alternative for predicting structures when experimental data is limited.

Purpose of the Study:

• To evaluate computational methods for predicting protein oligomer structures in the CASP14 experiment.
• To assess the impact of manual refinement and domain splitting on oligomer prediction accuracy.
• To compare docking strategies using AlphaFold2 monomers against traditional methods.

Main Methods:

• Utilized a server and human-expert approaches for oligomer structure prediction.
• Employed ab initio docking, automated refinement, and literature-based modeling.
• Performed post-prediction docking with AlphaFold2 monomers as input.

Main Results:

• Manual domain splitting and literature-based modeling improved oligomer structure prediction.
• Automated refinement enhanced subunit orientation and terminal structure accuracy.
• Ab initio docking of AlphaFold2 monomers did not yield superior oligomer predictions compared to blind predictions.

Conclusions:

• Manual intervention and integration of domain-specific information are vital for accurate oligomer modeling.
• AlphaFold2 monomer structures may not directly translate to improved oligomer predictions due to interfacial differences.
• Future research requires novel approaches, potentially incorporating evolutionary data or flexible docking, for advanced protein assembly prediction.