Predicting Anti-inflammatory Peptides by Ensemble Machine Learning and Deep Learning.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Pharmacology

Background:

• Inflammation is crucial for tissue repair but can lead to pathology if excessive or chronic.
• Current anti-inflammatory treatments (NSAIDs, corticosteroids, immunosuppressants) have limitations including side effects and resistance.
• Anti-inflammatory peptides (AIPs) present a promising therapeutic avenue for managing inflammation.

Purpose of the Study:

• To develop and validate an advanced machine learning framework for accurate prediction of anti-inflammatory peptides (AIPs).
• To accelerate the discovery and investigation of novel AIPs for therapeutic applications.
• To provide insights into AIP feature interpretation for future drug design.

Main Methods:

• Ensemble machine learning and deep learning framework integrating extremely randomized trees (ET), gated recurrent unit (GRU), and convolutional neural networks (CNNs) with attention.
• Utilized diverse sequence encodings and a stacking architecture to combine individual model strengths.
• Validated performance on an independent test set.

Main Results:

• The proposed ensemble model achieved high performance metrics: 0.757 accuracy, 0.500 MCC, and 0.707 F1-score on the independent test set.
• Demonstrated superior performance compared to contemporary anti-inflammatory peptide prediction methods.
• Provided valuable feature interpretation insights for AIPs.

Conclusions:

• The developed machine learning framework effectively predicts anti-inflammatory peptides (AIPs).
• This approach significantly advances the discovery and development of novel anti-inflammatory therapeutics.
• The study lays a foundation for designing targeted anti-inflammatory strategies.