Feedback-AVPGAN: Feedback-guided generative adversarial network for generating antiviral peptides.

Area of Science:

• Computational biology
• Peptide science
• Drug discovery

Background:

• Antiviral peptides (AVPs) are crucial for combating viral infections.
• Generating novel AVPs computationally is challenging due to limited experimental data.
• Existing methods struggle with the scarcity of experimentally validated AVPs.

Purpose of the Study:

• To develop a computational system, Feedback-AVPGAN, for generating novel antiviral peptides (AVPs).
• To address the data scarcity issue in training generative models for AVPs.
• To efficiently identify potential AVP candidates with high antiviral activity probability.

Main Methods:

• Utilized a Generative Adversarial Network (GAN) framework with a generator and discriminator.
• Implemented a Feedback method enabling the discriminator to learn from both real and synthetic data.
• Employed a transformer network-based classifier for high-accuracy peptide classification.
• Modeled generated peptide structures using AlphaFold2 for property and structure analysis.

Main Results:

• Successfully generated novel peptide sequences with a high probability of antiviral activity.
• The Feedback method enhanced the discriminator's learning from limited experimental AVP data.
• Generated peptides exhibited physicochemical properties and structures similar to known AVPs.
• AlphaFold2 modeling confirmed structural relevance of the computationally designed peptides.

Conclusions:

• Feedback-AVPGAN offers an effective computational approach for discovering novel antiviral peptides.
• The integration of feedback mechanisms overcomes data limitations in generative modeling for drug discovery.
• This system holds promise for accelerating the development of new antiviral therapeutics.