DeepAIPs-SFLA: Deep Convolutional Model for Prediction of Anti-Inflammatory Peptides Using Binary Pattern...

Area of Science:

• Computational biology
• Immunology
• Drug discovery

Background:

• Inflammation is a critical immune response, but chronic inflammation causes severe diseases.
• Anti-inflammatory peptides (AIPs) show therapeutic promise due to selectivity and minimal side effects.
• Existing computational AIP predictors lack internal sequence, structural, and evolutionary information.

Purpose of the Study:

• To develop a novel deep learning model, DeepAIPs-SFLA, for accurate prediction of anti-inflammatory peptides.
• To integrate evolutionary and structural peptide features using advanced image-based encoding.
• To improve upon existing computational methods for identifying potential anti-inflammatory peptide therapeutics.

Main Methods:

• Transformed peptide sequences into 2D structural (RECM) and evolutionary (PSSM) images.
• Utilized Local Binary Patterns (LBP) and Completed Local Binary Patterns (CLBP) for texture descriptors.
• Integrated features using differential evolution and selected optimal features with a shuffled frog-leaping algorithm (SFLA) for deep residual convolutional neural network (RCNN) training.

Main Results:

• The DeepAIPs-SFLA model achieved 97.04% predictive accuracy and an AUC of 0.98 on training sequences.
• Demonstrated significant accuracy improvements of 13% and 2% on independent datasets (Ind-426 and Ind-1049) compared to existing predictors.
• Validated the model's robustness and generalization power for predicting anti-inflammatory peptides.

Conclusions:

• DeepAIPs-SFLA offers a powerful computational approach for identifying anti-inflammatory peptides.
• The model's integration of diverse sequence and structural information enhances predictive accuracy.
• This tool holds potential for advancing research and accelerating drug discovery for inflammatory conditions.