Related Experiment Video
Updated: Aug 10, 2026
06:50
Computational Prediction of Amino Acid Preferences of Potentially Multispecific Peptide-Binding Domains Involved in Protein-Protein Interactions
Published on: January 26, 2024
DeepACPred: an integrated multistage framework for anticancer peptide discovery and activity prediction
Bo Zhang1,2, Ruifang Li1,3, Kedong Yin2,4
1Key Laboratory of Functional Molecules for Biomedicine of Zhengzhou City, Henan University of Technology, Zhengzhou, Henan 450001, P. R. China.
Briefings in Bioinformatics
|August 8, 2026
Summary
DeepACPred accelerates anticancer peptide (ACP) discovery using AI. This new framework integrates identification and activity prediction, successfully identifying novel ACP candidates with in vitro cytotoxic activity.
Area of Science:
- Computational Biology
- Drug Discovery
- Bioinformatics
Background:
- Artificial intelligence (AI) is crucial for accelerating the discovery of anticancer peptides (ACPs).
- Existing computational methods often lack integrated identification and activity-based prioritization of ACP candidates.
- There is a need for advanced computational frameworks to streamline ACP discovery and development.
Purpose of the Study:
- To present DeepACPred, a novel three-stage computational pipeline for ACP discovery and prioritization.
- To integrate ACP identification, cancer-type prediction, and activity (IC50) prediction using multimodal features.
- To validate the performance and applicability of DeepACPred in identifying potent ACPs.
Main Methods:
- Developed a three-stage pipeline: ACP binary classification, ACP multilabel classification (cancer type), and ACP IC50 prediction.
- Utilized multimodal features including ESM2 protein language model embeddings, AAindex physicochemical descriptors, and sequence composition.
- Evaluated models on benchmark datasets and applied DeepACPred to a large set of candidate peptides.
Main Results:
- The binary classifier achieved 95.10% accuracy (AUC=0.9913) for ACP identification.
- Multilabel cancer-type prediction yielded a macro-F1 score of 0.9124 across seven cancer types.
- IC50 prediction achieved a Spearman correlation of 0.8602, and 12 selected peptides showed in vitro cytotoxic activity (IC50: 0.88-36.83 μg/ml).
Conclusions:
- DeepACPred offers a systematic framework for enhancing and prioritizing anticancer peptide candidates.
- The pipeline effectively integrates multiple prediction tasks, leveraging multimodal features for improved performance.
- Experimental validation confirmed the in vitro activity of DeepACPred-identified peptides, supporting its utility in drug discovery.
