StackedEnC-AOP: prediction of antioxidant proteins using transform evolutionary and sequential features based...

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Antioxidant proteins protect cells and DNA from free radical damage.
• They play a key role in oxidative stress regulation and are targets for drug development.
• Current in vitro methods for identifying antioxidant proteins are inefficient and costly.

Purpose of the Study:

• To develop an accurate computational method for predicting antioxidant proteins.
• To overcome the limitations of current in vitro screening techniques.
• To provide a valuable tool for researchers and drug designers.

Main Methods:

• Proposed StackedEnC-AOP, a prediction method using Pseudo position-specific scoring matrix with discrete wavelet transform (PsePSSM-DWT).
• Integrated evolutionary difference and physiochemical properties for comprehensive feature extraction.
• Employed Minimum Redundancy Maximum Relevance (mRMR) for feature selection and a stacking-based ensemble meta-model for training.

Main Results:

• The StackedEnC-AOP model achieved 98.40% accuracy and 0.99 AUC on training sequences.
• Independent set validation demonstrated 96.92% accuracy and 0.98 AUC.
• The method significantly outperformed existing computational models.

Conclusions:

• StackedEnC-AOP offers a highly accurate and consistent approach to antioxidant protein prediction.
• The model shows substantial improvements in accuracy over current computational methods.
• This tool is valuable for data scientists, academia, and accelerating drug design processes.