PepMCP

a peptide-specific MCP predictor using GraphSAGE

PepMCP server

Requirements

ESMC environment

• python == 3.12
• pytorch == 2.7
• esm
• transformers

PepMCP environment

• pytorch
• transformers
• dgl
• sklearn (for training evaluation)

📌 Note: CPU-only ESMC.yml and PepMCP.yml files are provided for reference.

Usage

Prediction

Test sequences and names are in ./data/test.csv

python ESMC_extract.py  # ESMC environment
python predict.py       # PepMCP environment

Training

Data processing

(This step can be skipped for using 5-fold datasets in ./data/res_split or ./data/seq_split directly)

python preprocess.py
python ESMC_extract.py   # change file name to MemAMPs.csv or pdb_sol_neg.txt, under ESMC environment

Run

python train.py  # PepMCP environment

Citation

@article{Dong2026PepMCP,
    author = {Dong, Ruihan and Awang, Tadsanee and Cao, Qiushi and Kang, Kai and Wang, Lei and Zhu, Zefeng and Song, Chen},
    title = {PepMCP: a graph-based membrane contact probability predictor for membrane-lytic antimicrobial peptides},
    journal = {Bioinformatics},
    volume = {42},
    number = {7},
    pages = {btag453},
    year = {2026},
    month = {07},
    issn = {1367-4811},
    doi = {10.1093/bioinformatics/btag453},
}