Prediction of enzyme function by combining sequence similarity and protein interactions

Area of Science:

• Bioinformatics
• Computational Biology
• Enzymology

Background:

• Protein interaction data is increasingly used for protein function prediction.
• Existing methods often rely solely on protein sequence similarity.
• A novel approach is needed to integrate interaction data for enhanced functional annotation.

Purpose of the Study:

• To develop and evaluate a computational method for enzyme annotation.
• To leverage protein interaction networks for improved functional prediction.
• To enhance the accuracy of enzyme classification using sequence and interaction data.

Main Methods:

• A computational approach integrating protein sequence and interaction data was developed.
• The method identifies enzymes based on the principle that similar sequences with similar interaction partners perform similar functions.
• Performance was benchmarked against PSI-BLAST using a dataset of 3,890 protein sequences.

Main Results:

• The proposed method demonstrated a significant increase in prediction specificity for enzymes.
• Specificity for predicting the first three EC digits rose from 80% to 90% at 80% coverage compared to PSI-BLAST.
• The method proved effective for proteins with detectable homologous sequences and known interaction partners.

Conclusions:

• The novel computational approach enhances enzyme function prediction accuracy.
• Integrating protein interaction data alongside sequence information offers a significant improvement over sequence-based methods alone.
• This method holds potential for improving genome-wide enzyme predictions and understanding protein function.