Performance of rotation forest ensemble classifier and feature extractor in predicting protein interactions using...

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Predicting protein-protein interactions (PPIs) faces challenges in sequence representation and model design.
• Effective feature extraction is crucial for enhancing PPI prediction model performance.

Purpose of the Study:

• To evaluate global encoding and PseudoSMR for representing amino acid sequences in PPI prediction.
• To compare principal component analysis (PCA) and independent principal component analysis (IPCA) for transforming Rotation Forest classifiers.

Main Methods:

• Utilized global encoding and PseudoSMR for amino acid sequence feature extraction.
• Employed Rotation Forest classifier with PCA and IPCA transformations.
• Applied methods to human proteins and Human Immunodeficiency Virus type 1 (HIV-1) datasets.

Main Results:

• Global encoding and PseudoSMR effectively represented amino acid sequences for PPI classification.
• Rotation Forest (PCA) and Rotation Forest (IPCA) achieved evaluation metrics >73% across six parameters.
• Accuracy for both methods exceeded 74%, with sensitivity, specificity, precision, and F1-score also >73%.

Conclusions:

• Global encoding and PseudoSMR are effective methods for amino acid sequence representation in PPI prediction.
• Rotation Forest (PCA) outperformed Rotation Forest (IPCA) for HIV-1 and human protein interactions.
• Rotation Forest classifiers (with PCA/IPCA) surpassed other models like SVM and Random Forest, achieving >70% performance metrics.