Extracting features from protein sequences to improve deep extreme learning machine for protein fold recognition

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Protein fold recognition is crucial for predicting protein 3D structures.
• Efficient feature extraction from amino-acid sequences is key for accurate classification.
• Existing methods face challenges in optimizing feature representation.

Purpose of the Study:

• To propose a novel three-stage framework for protein fold recognition.
• To enhance feature extraction and classification performance using machine learning.
• To improve the prediction accuracy of protein three-dimensional structures.

Main Methods:

• Proposed a three-stage framework: PCA-DELM-LDA.
• Utilized six descriptors for initial feature extraction from protein sequences.
• Applied Principal Component Analysis (PCA) for feature reduction.
• Employed Deep Extreme Learning Machine (DELM) for feature enhancement.
• Used Linear Discriminant Analysis (LDA) for final classification into 27 folds.
• Implemented the framework on SCOP datasets with independent and combined feature sets.

Main Results:

• The proposed framework demonstrated improved performance in protein fold recognition.
• Feature vectors extracted in the first stage enhanced DELM's performance in the second stage.
• The framework successfully classified protein instances into 27 distinct folds.
• Experimental results validated the effectiveness of the proposed feature extraction and classification approach.

Conclusions:

• The novel three-stage PCA-DELM-LDA framework effectively improves protein fold recognition.
• Feature extraction and enhancement are critical for boosting the performance of protein structure prediction models.
• The proposed method offers a promising approach for advancing bioinformatics and structural biology research.