Welcome to PepNet!

        PepNet is a sequence-based deep learning model for the prediction of both AIPs and AMPs

        We provides a convenient and efficient PepNet webserver for researchers to predict the antimicrobial or anti-inflammatory activity of peptide sequences.

Fast mode:

Standard mode:

  1. Create an environment
    • # create
      • $ conda create -n PepNet python=3.8
    • # activate
      • $ source activate PepNet
  2. Install PepNet dependencies
    • (1) Install pytorch 2.4.0 (For more details, please refer to https://pytorch.org/)
      • For linux:
      # CUDA 11.8
      $ pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
    • (2) Install the requirements.
      • For linux:
      $ pip install h5py==3.11.0
      $ pip install transformers==4.43.2
    • (3) Extraction the pretrained features generated by ProtT5-XL-U50
      • (1) Download the source code of ProtTrans from https://github.com/agemagician/ProtTrans.
      • (2) Download the ProtT5-XL-U50 model from https://huggingface.co/Rostlab/prot_t5_xl_uniref50/tree/main.
      • (3) Place the downloaded model files in this directory {ProtTrans_path}/Embedding/Rostlab/prot_t5_xl_half_uniref50-enc
      • (4) Generate the pretrained feature for a fasta file.
        • $ cd {ProtTrans_path}/Embedding
        $ python prott5_embedder.py --input {the path of the fasta file} --output {the path of the output H5PY file, e.g, out.h5} --model {the path of the ProtT5-XL-U50 model you download}
  3. Standard mode using

If you think PepNet is useful, please kindly cite the following paper:

PepNet: a sequence-based deep learning model for predicting anti-inflammatory and antimicrobial peptides

Webserver update:

May, 13th, 2024: the first version of PepNet server was established.

This work is openly licensed via CC0 1.0