predCar-site: Carbonylation sites prediction in proteins using support vector machine with resolving data imbalanced...

Area of Science:

• Biochemistry
• Computational Biology
• Proteomics

Background:

• Protein carbonylation is an irreversible post-translational modification and a biomarker for oxidative stress.
• It is implicated in various biological processes and diseases, including Alzheimer's, diabetes, and Parkinson's.
• Experimental detection of carbonylation sites is costly and time-consuming, necessitating computational approaches.

Purpose of the Study:

• To develop a novel computational tool, predCar-Site, for accurate prediction of protein carbonylation sites.
• To address the urgent need for efficient methods in drug development related to carbonylation-associated diseases.

Main Methods:

• Incorporated sequence-coupled information into general pseudo amino acid composition.
• Balanced skewed training datasets using the Different Error Costs method.
• Employed a support vector machine classifier for prediction.

Main Results:

• The predCar-Site predictor achieved high average AUC scores: 0.9959 for K, 0.9999 for P, 1 for R, and 0.9997 for T.
• Results were validated using 10-fold cross-validation over 5 runs.
• Demonstrated significantly superior performance compared to existing prediction tools.

Conclusions:

• predCar-Site offers a highly accurate and efficient method for predicting protein carbonylation sites.
• The tool can significantly aid in understanding oxidative stress mechanisms and accelerate drug discovery for related diseases.
• A user-friendly web server is available for public access.