An aggregate analysis of many predicted structures to reduce errors in protein structure comparison caused by...

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Science

Background:

• Conformational flexibility in proteins leads to inaccuracies in structure comparison.
• Altered ligand binding cavities due to conformational changes can misclassify protein function and evolutionary relationships.

Purpose of the Study:

• To apply protein structure prediction algorithms to improve the classification of homologous proteins based on binding preferences.
• To address challenges in comparing protein structures with significant conformational differences.

Main Methods:

• Utilized protein structure prediction to "remodel" existing protein structures into comparable states.
• Introduced "medial remodeling" to mitigate errors by analyzing multiple predicted structures and eliminating extreme models of binding cavities.

Main Results:

• Remodeling enabled objective comparison of proteins with diverse conformations, improving classification accuracy for binding preferences.
• Medial remodeling successfully reduced errors caused by conformational perturbations, particularly in the enolase and tyrosine kinase superfamilies.
• Classification accuracy increased for proteins erroneously predicted to have different binding preferences.

Conclusions:

• Protein structure prediction, combined with medial remodeling, effectively compensates for conformational variability in comparing protein-ligand binding sites.
• This approach enhances the accuracy of protein structure comparison beyond existing crystal structures.
• Medial remodeling provides a robust method to handle uncertainties introduced by structure prediction.