ConsAMPHemo: A computational framework for predicting hemolysis of antimicrobial peptides based on machine learning
Peilin Xie1,2, Lantian Yao1,2, Jiahui Guan3
1Kobilka Institute of Innovative Drug Discovery, School of Medicine, The Chinese University of Hong Kong, Shenzhen, China.
Protein Science : a Publication of the Protein Society
|June 16, 2025
Summary
This study introduces ConsAMPHemo, a deep learning framework to predict antimicrobial peptide (AMP) hemolysis. It accurately classifies hemolytic activity and predicts hemolysis concentrations, aiding in the development of safer AMP drugs.
Area of Science:
- Biochemistry and Molecular Biology
- Computational Biology and Bioinformatics
- Drug Discovery and Development
Background:
- Antimicrobial peptides (AMPs) are crucial for fighting microbes by disrupting cell membranes.
- The membrane-disrupting mechanism of AMPs can cause unintended hemolysis, posing a safety concern for therapeutic applications.
- Current experimental methods for assessing AMP hemolytic activity are costly and time-consuming.
Purpose of the Study:
- To develop a cost-effective, deep learning-based computational framework for predicting the hemolytic activity of AMPs.
- To classify AMPs as hemolytic or non-hemolytic and predict their specific hemolysis concentrations.
- To identify key features correlating with hemolytic activity and understand the biophysical basis of AMP-induced hemolysis.
Main Methods:
- Development of ConsAMPHemo, a two-stage deep learning framework.
- Implementation of binary classification for hemolytic activity prediction.
- Application of regression analysis for predicting hemolysis concentrations.
Main Results:
- ConsAMPHemo achieved high classification accuracy (99.54%, 82.57%, 88.04%) across three datasets.
- The model demonstrated strong performance in regression prediction with a Pearson correlation coefficient of 0.809.
- Identified significant correlations between specific peptide features and hemolytic activity.
Conclusions:
- ConsAMPHemo offers a computationally efficient and accurate method for predicting AMP hemolytic activity, reducing the need for extensive experimental testing.
- The framework aids in the rational design of safer AMPs with minimized hemolytic toxicity.
- Insights into the physics of hemolysis facilitate the development of next-generation antimicrobial therapeutics.
