PepAnno: A structure-aware deep learning framework for bioactive peptide prediction, structural visualization, and
Enyan Liu1, Yueming Hu1, Liya Liu1
1Department of Bioinformatics, College of Life Sciences, Zhejiang University, Hangzhou, China.
Plos Computational Biology
|June 2, 2026
Summary
PepAnno is a new web server for peptide annotation, offering a user-friendly interface and structure-aware deep learning for predicting bioactivities. It improves research efficiency for peptide drug discovery.
Area of Science:
- Computational biology
- Drug discovery
- Bioinformatics
Background:
- Peptides are promising therapeutic agents, but existing prediction tools have limitations.
- Current methods often lack structural awareness, interpretability, and efficient workflows.
Purpose of the Study:
- To develop a comprehensive, user-friendly web server for multi-functional peptide annotation.
- To address limitations of existing tools by integrating structural and sequence information.
Main Methods:
- Developed PepAnno, a web server utilizing a structure-aware, multi-view geometric deep learning framework.
- Integrated pre-trained sequence embeddings with predicted 3D structural graphs using a dual-stream Transformer and GATv2 architecture.
- Employed a cross-modal attention mechanism for fusing semantic and geometric representations.
Main Results:
- PepAnno accurately predicts 7 key peptide bioactivities, including antimicrobial and anticancer properties.
- Demonstrated robust and competitive performance, outperforming or matching existing methods.
- Provides automated physicochemical property calculation, structure visualization, and access to integrated databases.
Conclusions:
- PepAnno offers an efficient and interpretable solution for large-scale peptide analysis.
- Facilitates downstream experimental design and accelerates peptide-based drug discovery.
- Enhances research efficiency and reduces costs in bioactive peptide identification.
