PandoraGAN: Generating antiviral peptides using Generative Adversarial Network

Abstract

The continuous increase in pathogenic viruses and the intensive laboratory research for development of novel antiviral therapies often poses challenge in terms of cost and time efficient drug design. This accelerates research for alternate drug candidates and contributes to recent rise in research of antiviral peptides against many of the viruses. With limited information regarding these peptides and their activity, modifying the existing peptide backbone or developing a novel peptide is very time consuming and a tedious process. Advanced deep learning approaches such as generative adversarial networks (GAN) can be helpful for wet lab scientist to screen potential antiviral candidates of interest and expedite the initial stage of peptide drug development. To our knowledge this is the first ever use of GAN models for antiviral peptides across the viral spectrum. In this study, we develop PandoraGAN that utilizes GAN to design bio active antiviral peptides. Available antiviral peptide data was manually curated for preparing highly active peptides data set to include peptides with lower IC50 values. We further validated the generated sequences comparing the physico-chemical properties of generated antiviral peptides with manually curated highly active training data. Antiviral sequences generated by PandoraGAN are available on PandoraGAN server.

Proteochemometric Method for pIC50 Prediction of Flaviviridae

“…Although elusive, antiviral peptides do exhibit certain physicochemical properties which makes them great candidates for antiviral therapeutics [ 1 ]. The work done by Surana et al explains the usage of physicochemical properties [ 36 ]. Although physicochemical properties do help in identifying the potential antiviral peptides, IC50 is one of the methods used for further validating the efficacy of candidate peptides [ 37 ].…”

Proteochemometric method for pIC50 prediction of Flaviviridae

“…Although elusive, antiviral peptides do exhibit certain physicochemical properties which makes them great candidates for antiviral therapeutics [37]. The work done by Surana et al explained the usage of physicochemical properties [38]. Although physicochemical properties do help in identifying the potential antiviral peptides, IC50 is one of the methods used for further validating the efficacy of candidate peptides [39].…”

Deep generative models for peptide design

“…However, some peptide generation studies have used representations that cannot be mapped directly to a unique sequence in property predictors to lter out those peptides generated that possess undesired properties. For instance, Rossetto et al 48 used a 4D tensor Schemes commonly used to represent peptides in the generation task itself include direct sequence representation 41,42,45,46,[49][50][51][52][53] and learned embeddings. 38,39,44,45,54,55 A natural way to encode peptides is through their primary structure (i.e., amino acid sequence).…”

“…38,39,44,45,54,55 A natural way to encode peptides is through their primary structure (i.e., amino acid sequence). A peptide of length L can be represented by a string of characters or integers of length L, 39,45,46,49,50 or a L Â n matrix such that each amino acid has a unique n-dimensional vector. The n-dimensional vector may represent either experimentally 51 or computationally derived properties, 45 or be a one-hot encoding.…”

“…Its improved version, AMP-GAN v2, exhibited better training stability and generated a higher percentage of peptide antibiotic candidates, as predicted by several machine learning-based AMP classifiers. PandoraGAN 49 utilized around 100–400 active antiviral peptides to generate novel peptides targeting viruses. To address the data scarcity problem, a novel training framework proposed in LeakGAN 100 was adopted to provide richer signals and guide the model training during the intermediate process of peptide generation.…”