AIP-TranLAC: A Transformer-Based Method Integrating LSTM and Attention Mechanism for Predicting Anti-inflammatory...

Area of Science:

• Biotechnology
• Bioinformatics
• Computational Biology

Background:

• Anti-inflammatory peptides (AIPs) show therapeutic potential for inflammatory diseases.
• Computational identification of AIPs is currently a significant challenge.

Purpose of the Study:

• To introduce AIP-TranLAC, a novel deep learning framework for accurate AIP classification.
• To improve the computational identification of therapeutic anti-inflammatory peptides.

Main Methods:

• A hybrid deep learning architecture combining Transformer-based embedding, Bi-LSTM, multi-head attention, and CNN.
• Utilizing sequence patterns for accurate classification of anti-inflammatory peptides.
• Performing interpretability analyses to identify critical amino acid residues.

Main Results:

• AIP-TranLAC achieved superior performance on benchmark and independent datasets.
• The model demonstrated significant improvements over existing computational methods.
• The framework effectively captured both local and global sequence patterns in peptides.

Conclusions:

• AIP-TranLAC offers a powerful and accurate tool for accelerating the discovery of therapeutic peptides.
• The model's interpretability aids in understanding the mechanisms of anti-inflammatory action.
• Open-source availability promotes reproducibility and further research in inflammation and peptide discovery.