Mining channel-regulated peptides from animal venom by integrating sequence semantics and structural information

Area of Science:

• Biochemistry
• Bioinformatics
• Pharmacology

Background:

• Channel-regulated peptides (CRPs) from animal venom are promising drug candidates for channel protein-related diseases.
• Traditional methods for CRP discovery are slow and labor-intensive.
• Existing computational methods for CRP identification are limited in scope and feature engineering.

Purpose of the Study:

• To develop a novel deep learning model, DeepCRPs, for systematic mining of CRPs from animal venom.
• To enhance the classification performance of CRPs by integrating sequence and structural information.
• To provide bio-explainable insights into CRP identification and categorization.

Main Methods:

• Developed DeepCRPs, a deep learning model based on graph neural networks.
• Integrated sequence semantic and structural information for CRP classification.
• Applied advanced interpretable techniques to analyze sequence and structural determinants.

Main Results:

• Achieved a classification accuracy of 0.92 for four CRPs, significantly outperforming baseline models (0.77-0.89).
• Identified key sequence and structural determinants contributing to CRP classification.
• Demonstrated the precision and interpretability of the DeepCRPs model.

Conclusions:

• DeepCRPs offers an accurate and bio-explainable computational approach for identifying and categorizing CRPs.
• The model facilitates the discovery and development of novel toxin peptides targeting channel proteins.
• The study contributes to advancing drug discovery for channelopathies.