DP-AOP

DP-AOP identifies antioxidant proteins by combining Wei's feature-extraction, MRMD scoring, dynamic programming-based subset optimization, SMOTE data balancing, and SVM classification implemented via libsvm.

Key Features:

• Machine Learning: DP-AOP uses a Support Vector Machine (SVM) classifier implemented with libsvm for antioxidant protein prediction.
• Data Balancing (SMOTE): The pipeline applies the Synthetic Minority Over-sampling Technique (SMOTE) to balance class representation in the training dataset.
• Feature Extraction: Wei's proposed algorithm generates 473-dimensional feature vectors used as the initial feature set.
• Feature Scoring (MRMD): MRMD's sorting function scores and ranks features by contribution value to prioritize informative attributes.
• Dynamic Programming-based Selection: Dynamic programming optimizes local eight-feature subsets, reducing dimensionality to a 36-dimensional set and selecting 17 optimal features for model training.
• Performance Metrics: Reported evaluation metrics are accuracy 91.076%, sensitivity (SN) 96.4%, specificity (SP) 85.8%, Matthews correlation coefficient (MCC) 82.6%, and F1 score 91.5%.

Scientific Applications:

• Oxidative Stress Research: Identification of antioxidant proteins to study mechanisms of protection against free-radical damage in biological systems.
• Protein Classification in Molecular Biology: Classification of antioxidant versus non-antioxidant proteins to support biochemistry and molecular biology investigations and aid discovery of therapeutic targets and understanding of cellular defense mechanisms.

Methodology:

Data balancing with SMOTE; feature extraction using Wei's algorithm to produce 473-dimensional vectors; MRMD scoring and ranking; dynamic programming optimization of local eight-feature subsets leading to a 36-dimensional set and selection of 17 features; SVM training and evaluation via libsvm with reported metrics.