Detection of Interactions between Proteins through Rotation Forest and Local Phase Quantization Descriptors

Area of Science:

• Computational Biology
• Bioinformatics
• Proteomics

Background:

• Protein-protein interactions (PPIs) are crucial for cellular functions.
• Experimental methods for detecting PPIs are costly and time-consuming.

Purpose of the Study:

• To develop a novel, efficient, and accurate computational method for predicting PPIs.
• To leverage protein sequence information for improved PPI prediction.

Main Methods:

• Utilized the Rotation Forest (RF) classifier.
• Employed the Local Phase Quantization (LPQ) descriptor derived from the Physicochemical Property Response (PR) Matrix.
• Applied the method to Saccharomyces cerevisiae, Homo sapiens, and Helicobacter pylori PPI datasets.

Main Results:

• Achieved high average accuracies: 93.8% for S. cerevisiae, 97.96% for H. sapiens, and 89.47% for H. pylori.
• Demonstrated superior performance compared to previous studies.
• Validated the Rotation Forest classifier against the Support Vector Machine (SVM) classifier.

Conclusions:

• The proposed computational method offers a highly efficient and accurate approach for PPI prediction.
• This method can serve as a valuable complement to experimental techniques in proteomics research.