Related Experiment Video
Updated: Jul 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
RLAnOxPeptide: an integrated framework combining transformer and reinforcement learning for efficient antioxidant
Changsheng Han1,2,3, Jianda Yue1,2,3, Yaqi Li1,2,3
1The National and Local Joint Engineering Laboratory of Animal Peptide Drug Development, College of Life Sciences, Hunan Normal University, Changsha, Hunan 410081, China.
Bioinformatics (Oxford, England)
|July 9, 2026
Summary
This study introduces RLAnOxPeptide, a computational framework for discovering antioxidant peptides (AOPs). It efficiently predicts and designs novel AOPs with validated radical scavenging and cellular protective effects.
Area of Science:
- Biotechnology
- Computational Biology
- Peptide Science
Background:
- Bioactive peptides, particularly antioxidant peptides (AOPs), hold significant promise in pharmaceuticals and food science.
- Traditional methods for discovering AOPs are inefficient and expensive.
- There is a need for advanced computational tools to accelerate AOP discovery.
Purpose of the Study:
- To develop an integrated computational framework, RLAnOxPeptide, for efficient prediction and de novo design of AOPs.
- To merge machine learning and reinforcement learning for enhanced AOP discovery.
- To validate the designed AOPs experimentally.
Main Methods:
- Developed RLP-T5Pred, a high-precision predictor using ProtT5 and a 'protein-to-peptide' knowledge transfer strategy.
- Implemented RLP-T5Gen, a generator trained with supervised and reinforcement learning in a 'Yin-Yang' loop.
- Utilized a multi-objective reward function and RLP-T5Pred as an evaluator for novel AOP design.
- Experimentally synthesized and validated 17 designed peptides using chemical and cellular assays.
Main Results:
- RLP-T5Pred achieved state-of-the-art accuracy (AUC-ROC: 0.9692) with robust calibration.
- RLP-T5Gen successfully designed novel AOPs with high predicted activity.
- 17 designed peptides showed potent radical scavenging abilities.
- Four peptides (Pep4, Pep5, Pep10, Pep11) demonstrated significant protective effects against oxidative damage in HepG2 cells.
Conclusions:
- The RLAnOxPeptide framework offers a powerful, experimentally verified paradigm for accelerating the discovery of novel antioxidant peptides.
- The integrated approach of machine learning and reinforcement learning is effective for de novo peptide design.
- The validated AOPs show potential for pharmaceutical and food applications.
