Random Forest classification based on star graph topological indices for antioxidant proteins.

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Aging and quality of life are increasingly important research areas.
• Antioxidant proteins show potential in influencing the aging process.
• Experimental testing of all proteins for antioxidant activity is costly and inefficient.

Purpose of the Study:

• To develop a computational model for predicting antioxidant proteins.
• To reduce the number of proteins requiring experimental validation.
• To leverage complex network graphs and topological indices for protein analysis.

Main Methods:

• Representing protein primary structures as complex network graphs.
• Utilizing Randić's Star Networks and associated topological indices.
• Calculating indices with the S2SNet tool on a dataset of 1999 proteins (324 antioxidant).
• Employing classification techniques, including Random Forest, for prediction.

Main Results:

• The Random Forest model achieved 94% overall classification accuracy.
• The model successfully identified antioxidant proteins, achieving 81.8% accuracy for this minority class.
• A precision of 81.3% was obtained for the antioxidant protein class.

Conclusions:

• The proposed graph-based model is effective for identifying antioxidant proteins.
• This approach offers a more efficient and cost-effective alternative to traditional experimental methods.
• The model demonstrates strong performance even with imbalanced datasets.