Radial basis function neural network ensemble for predicting protein-protein interaction sites in heterocomplexes

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Protein-protein interactions are crucial for cellular functions.
• Identifying interaction sites is key to understanding protein complex structures.
• Existing methods for predicting interaction sites have limitations.

Purpose of the Study:

• To develop and validate a novel ensemble model for predicting protein-protein interaction sites in heterocomplexes.
• To utilize evolutionary conservation and spatial sequence profiles for accurate prediction.
• To improve the efficiency and effectiveness of interaction site prediction.

Main Methods:

• A radial basis function neural network (RBFNN) ensemble model was employed.
• Protein surface residues were classified as interaction or non-interaction sites.
• The model was trained and validated on a non-redundant dataset of 69 heterodimer protein chains.

Main Results:

• The RBFNN ensemble model achieved an accuracy of 0.689.
• The model demonstrated a sensitivity of 66.6% and a specificity of 67.6%.
• The predictor effectively uses evolutionary conservation and spatial sequence profiles.

Conclusions:

• The proposed RBFNN ensemble method is effective for predicting protein-protein interaction sites.
• This approach aids in the structural elucidation of protein complexes.
• The method shows promising performance metrics for interaction site prediction.