Protein function prediction using neighbor relativity in protein-protein interaction network

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• A significant number of discovered proteins lack functional annotation.
• Experimental determination of protein function is costly and time-consuming.
• Computational methods, particularly those using protein-protein interaction networks, are crucial for function prediction.

Purpose of the Study:

• To propose a novel computational approach, the Neighbor Relativity Coefficient (NRC), for predicting protein function.
• To estimate functional similarity between proteins based on interaction network topology.
• To improve the accuracy of protein function prediction for un-annotated proteins.

Main Methods:

• Developed the Neighbor Relativity Coefficient (NRC) based on graph-theoretic features (distance, common neighbors, paths).
• Utilized NRC to weight neighbor annotations for transferring function to un-annotated proteins.
• Evaluated the NRC method on protein-protein interaction networks of Saccharomyces cerevisiae and Homo sapiens.

Main Results:

• The NRC method effectively estimates functional similarity between proteins.
• NRC successfully transfers annotations to un-annotated proteins based on network topology.
• Performance analysis showed NRC outperforms previous interaction network-based function prediction methods.

Conclusions:

• The Neighbor Relativity Coefficient (NRC) offers a robust and effective method for protein function prediction.
• NRC enhances the understanding of protein roles by leveraging interaction network structures.
• This approach addresses the challenge of functional annotation gaps in bioinformatics.