PandoraGAN

PandoraGAN generates novel antiviral peptide (AVP) sequences using a Generative Adversarial Network (GAN) to produce candidates that recapitulate physico-chemical properties of known highly active AVPs for accelerated antiviral peptide discovery.

Key Features:

• Generative Adversarial Network (GAN) architecture: Uses a GAN to learn sequence and property distributions from known antiviral peptides.
• Manually curated training dataset: Trained on a dataset of 130 highly active peptides curated from AVPdb and relevant literature.
• Physico-chemical property preservation: Generates sequences that exhibit similar physico-chemical properties to the training AVPs.
• Statistical validation: Performs statistical comparisons between generated sequences and the training set using Pearson’s correlation and the Mann-Whitney U-test.
• Implicit property learning: Model architecture is designed to capture and reproduce implicit properties inherent to antiviral peptides.
• Novel sequence generation: Produces candidate peptide sequences that can reveal patterns not present in the original dataset.

Scientific Applications:

• Antiviral peptide candidate generation: Produces novel AVP sequences for downstream experimental evaluation and lead selection.
• Sequence space expansion: Broadens the searchable AVP sequence space to identify potential candidates against diverse viral pathogens.
• Comparative property analysis: Facilitates comparison of physico-chemical properties between generated candidates and known highly active AVPs.
• Computational lead discovery: Supports early-stage computational discovery in antiviral peptide drug development workflows.

Methodology:

Train a Generative Adversarial Network on a manually curated set of 130 highly active peptides sourced from AVPdb and literature; generate novel peptide sequences with the GAN architecture engineered to capture implicit AVP properties; validate generated sequences by statistical comparison to the training dataset using Pearson’s correlation and the Mann-Whitney U-test.