Accelerating Prediction of Antiviral Peptides Using Genetic Algorithm-Based Weighted Multiperspective Descriptors...

Area of Science:

• Computational Biology
• Bioinformatics
• Drug Discovery

Background:

• Viral diseases pose a significant global health challenge despite existing antiviral medications.
• Antiviral peptides (AVPs) show promise as novel therapeutic agents.
• Traditional methods for identifying AVPs are inefficient and lack deep sequence insights.

Purpose of the Study:

• To develop a precise and efficient computational model for predicting antiviral peptides (AVPs).
• To overcome the limitations of traditional labor-intensive and expensive AVP identification methods.
• To enhance the understanding of peptide mechanisms in antiviral drug development.

Main Methods:

• Utilized ProtGPT2 for contextual peptide embeddings and sequence-to-image transformations (SMR, RECM).
• Applied CLBP for local image decomposition and differential evolution for feature vector formation.
• Employed a hybrid MRMD + SFLA approach for optimal feature selection.
• Developed a novel self-normalized capsule network (Sn-CapsNet) for prediction.

Main Results:

• Achieved a superior predictive accuracy of 97.36% for AVPs.
• Outperformed existing predictors by approximately 12% with an AUC of 0.98.
• Demonstrated robust generalization on an independent dataset with an 8% improvement over previous models.

Conclusions:

• The TargetAVP-DeepCaps model offers a highly accurate and efficient tool for AVP prediction.
• This computational approach accelerates the discovery and development of peptide-based antiviral therapeutics.
• Provides a valuable resource for understanding peptide mechanisms and their applications in drug discovery.