DeepAIP: Deep learning for anti-inflammatory peptide prediction using pre-trained protein language model features...

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Current anti-inflammatory treatments like NSAIDs and glucocorticoids have significant side effects.
• There is a growing need for safer and more effective anti-inflammatory therapies.
• Anti-inflammatory peptides (AIPs) represent a promising area for novel therapeutic development.

Purpose of the Study:

• To develop a novel deep learning model for accurate prediction of anti-inflammatory peptides (AIPs).
• To leverage contextual self-attention mechanisms and pre-trained protein language models for enhanced feature extraction.
• To evaluate and compare the performance of different protein language models in predicting AIPs.

Main Methods:

• A contextual self-attention deep learning model, DeepAIP, was proposed.
• Features were extracted using pre-trained protein language models, with Prot-T5 showing superior performance.
• The model's predictive accuracy was assessed on benchmark datasets and novel peptide sequences.

Main Results:

• DeepAIP achieved higher Matthews Correlation Coefficient (MCC) and Accuracy scores compared to existing methods on a benchmark dataset.
• Prot-T5 features provided the best comprehensive performance for the deep learning model.
• DeepAIP accurately identified all 17 novel anti-inflammatory peptide sequences in a performance comparison analysis.

Conclusions:

• The proposed DeepAIP model, utilizing contextual self-attention and Prot-T5 features, demonstrates high accuracy in predicting anti-inflammatory peptides.
• This deep learning approach offers a viable and effective strategy for identifying novel AIPs.
• DeepAIP shows potential for advancing the development of safer and more effective anti-inflammatory treatments.