Deep Learning-Based Bioactive Therapeutic Peptide Generation and Screening.

Area of Science:

• Biotechnology
• Computational Biology
• Drug Discovery

Background:

• Bioactive peptides show therapeutic potential for various diseases.
• Peptide synthesis is more cost-effective than traditional compound synthesis.
• Deep learning offers a novel approach for generating de novo peptides.

Purpose of the Study:

• To develop and apply deep learning models for generating de novo bioactive peptides.
• To create a pipeline for screening generated peptides against specific therapeutic targets.
• To demonstrate the iterative optimization of peptide generation for enhanced binding affinity.

Main Methods:

• Developed an LSTM-based model (LSTM_Pep) for de novo peptide generation.
• Utilized the Antimicrobial Peptide Database for training the generative model.
• Created a deep learning-based prediction model (DeepPep) for rapid screening.
• Applied a pipeline using LSTM_Pep and DeepPep for target-specific peptide discovery.

Main Results:

• Successfully generated diverse de novo peptides with potential bioactivity.
• Demonstrated the pipeline's efficacy using SARS-COV-2 main protease as a proof-of-concept.
• Achieved iterative fine-tuning for generating peptides with higher predicted binding affinity.
• Showcased AI's capability in discovering de novo bioactive peptides for specific targets.

Conclusions:

• Deep learning models can effectively generate bioactive peptides with desired therapeutic effects.
• The proposed pipeline facilitates efficient discovery and optimization of therapeutic peptides.
• AI-driven approaches significantly advance the field of de novo drug discovery for peptides.