Using Random Forest Model Combined With Gabor Feature to Predict Protein-Protein Interaction From Protein Sequence

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Protein-protein interactions (PPIs) are fundamental to cellular processes.
• Experimental PPI detection is expensive and time-consuming.
• Computational prediction of PPIs is essential due to data volume and experimental limitations.

Purpose of the Study:

• To develop a novel computational method for predicting PPIs.
• To utilize texture features of protein sequences for improved prediction accuracy.
• To establish a reliable and robust method for inferring protein interactions.

Main Methods:

• Extraction of texture features and evolutionary information using Gabor filters on Position-Specific Scoring Matrices (PSSMs).
• Generation of PSSMs using Position-Specific Iterated Basic Local Alignment Search Tool (PSI-BLAST).
• Classification of protein interactions using random forest algorithms.

Main Results:

• Achieved high prediction accuracies: 92.10% for yeast, 97.03% for human, and 86.45% for H. pylori.
• Demonstrated the reliability of Gabor features compared to Discrete Cosine Transform and Local Phase Quantization.
• Validated the method's feasibility, stability, power, and robustness across multiple species.

Conclusions:

• The proposed Gabor feature-based method is effective for computational PPI prediction.
• Gabor features are reliable for extracting relevant information from protein sequences.
• This method provides a powerful and robust tool for advancing PPI research.