Pore-Forming_AMP_SVM

Codes for Mechanism-Driven Screening of Membrane-Targeting and Pore-Forming Antimicrobial Peptides

Requirements

SVM

• thundersvm
• transformers (for prot_bert_bfd)
• torch
• sklearn

MCP and CM predictors

• numpy
• matplotlib
• pickleshare
• tensorflow==1.7.0 (if you are using tensorflow 2, please modify the script with 'import tensorflow.compat.v1 as tf')

Usage for SVM

For independent training/validation

python train_val.py

For 10-fold cross validation

python train_10fold.py

For predicting/screening

python predict.py

• Note: The length of input peptide should not be longer than 40 amino acids.

Usage for MCP and CM predictors

Requirements

pip install -r requirements.txt

Feature generation

• Input features are generated from SPIDER3-Single (Heffernan, R. et al., J Comput Chem 2018). Download

MCP prediction

python MCPpep_predictor.py ./example test

CM prediction

python CMpep_predictor.py ./example test

• Note: The complete parameter files can be downloaded at here.

Inter-chain CM plot

python plot_dimer.py test

Citation

@article{Li2025AMP,
  author = {Li, Jiaxuan and Yang, Chenguang and Dong, Ruihan and Juarez, Juan Francisco Bada and Wang, Lei and Wettstein, Maximilian Emanuel and Wang, Dali and Cao, Chan and Lu, Ying and Song, Chen},
  title = {Mechanism-Driven Screening of Membrane-Targeting and Pore-Forming Antimicrobial Peptides},
  journal = {Advanced Science},
  pages = {e16470},
  doi = {https://doi.org/10.1002/advs.202516470},
  year = {2025}
}