Related Experiment Video
Updated: Jan 6, 2026
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
1.6K
ACP-EPC: an interpretable deep learning framework for anticancer peptide prediction utilizing pre-trained protein
Jingwei Lv1, Kexin Li1, Yike Wang1
1School of Computer Science and Technology, Hainan University, Haikou, 570228, China.
Molecular Diversity
|September 13, 2025
Summary
Researchers developed ACP-EPC, a deep learning tool to predict anticancer peptides (ACPs) from protein sequences. This AI approach accelerates the discovery of novel ACPs, offering a promising alternative to traditional chemotherapy with fewer side effects.
Area of Science:
- Biochemistry
- Bioinformatics
- Computational Biology
Background:
- Conventional chemotherapy causes significant damage to healthy cells, leading to severe side effects.
- Anticancer peptides (ACPs) offer a targeted therapeutic alternative, selectively eliminating cancer cells.
- Experimental identification of ACPs is time-consuming and labor-intensive.
Purpose of the Study:
- To develop a deep learning framework, ACP-EPC, for accurate prediction of ACPs directly from protein sequences.
- To overcome the limitations of traditional experimental methods for ACP identification.
- To provide a publicly accessible tool for researchers in the field.
Main Methods:
- Developed ACP-EPC, a deep learning framework utilizing Evolutionary Scale Modeling 2 (ESM-2) for contextual representations.
- Integrated handcrafted physicochemical descriptors with ESM-2 features.
- Employed a Cross-Attention mechanism for multimodal feature fusion.
- Evaluated model performance using tenfold cross-validation and two independent test sets (ACP135 and ACP99).
Main Results:
- ACP-EPC achieved high prediction accuracy: 0.935 on tenfold cross-validation and 0.984 on the ACP99 test set.
- The model demonstrated superior performance compared to existing ACP prediction methods.
- The integration of diverse feature representations proved advantageous for prediction accuracy.
Conclusions:
- ACP-EPC is an effective deep learning framework for predicting anticancer peptides from protein sequences.
- The study highlights the benefits of combining evolutionary and physicochemical features for enhanced prediction.
- A publicly available web server (http://www.bioai-lab.com/ACP-EPC) has been established to facilitate ACP research.
