Machine learning-guided discovery and design of non-hemolytic peptides
Problem
Antimicrobial peptides are promising against antibiotic-resistant infections but their hemolytic toxicity blocks clinical use, and reliable computational prediction is hard.
Why it matters
Antibiotic-resistant infections need new therapies, and hemolytic toxicity is a key hurdle to clinical trials.
What the AI did
Developed machine learning models plus outlier detection methods to predict peptide activity and design novel peptides with reduced hemolytic activity.
Results
not stated in source
Impact
Evidence supports development of predictive models for non-hemolytic peptide design, but performance figures are cut off.
Record
| Venue | Scientific Reports |
|---|---|
| Published | 2020 |
| Work type | article |
| Authors | Fabien Plisson, Obed Ramírez‐Sánchez, Cristina Martínez-Hernández |
| Citations | 148 (OpenAlex, at import time) |
| References | 86 |
| Model or method | Gradient boosting (named as best models); machine learning with outlier detection |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41598-020-73644-6 |
| Open access | gold |
| Keywords | machine learning, antimicrobial peptides, hemolysis, drug design, outlier detection |
Importance rating
significant — thousands to millions affected
Addresses antibiotic-resistant infections, a major health threat, by removing hemolytic toxicity hurdles to clinical trials; abstract gives model accuracy but not disease burden scale.