EpitopeVec: Linear Epitope Prediction Using DeepProtein Sequence Embeddings

Contents

• EpitopeVec Introduction
• Requirements
• Usage
• Input
• Output
• License
• Citation
• Contact

EpitopeVec

EpitopeVec predicts linear B-cell epitopes. It takes protein sequences (FASTA format) as input and then lists the peptides (of a given amino acid length) that can be epitopes along with their respective prediction probability. It is based on a SVM model trained on a large set of experimentally verified epitopes and makes use of different amino acid features, amino acid pair propensity scale, amino acid trimer propensity scale, and distributed vector representation of proteins (ProtVec). The code for training new models and results from the EpitopVec article are available at https://github.com/hzi-bifo/epitope-prediction-paper

Requirements

• Python 3 with the following packages:

  • numpy 1.17.1
  • scipy 1.4.1
  • matplotlib 3.1.3
  • sklearn 0.22.1
  • pydpi 1.0
  • biopython 1.71.0
  • tqdm 4.15.0
  • gensim 3.8.3

  If these are not installed, you can install them with pip.

  pip3 install -r ./requirement/requirements.txt
  

  Additionally, pydpi 1.0 from pip might be incompatible with Python 3. Please install the pydpi package from the provided requirement/pydpi.tar.gz file.

  pip3 install ./requirement/pydpi-1.0.tar.gz
  

• Binary file for ProtVec representation of proteins can be downloaded using the following command in the protvec directory:

cd protvec
wget https://research.bifo.helmholtz-hzi.de/downloads/epitopevec/sp_sequences_4mers_vec.bin

If the link to the protvec binary file is broken or the download is too slow, please download the file (only the .bin file is required) from here and place it in the protvec directory.

Usage

• Clone this repository:

  git clone https://github.com/hzi-bifo/epitope-prediction
  

• Run the main.py file with the following command.

    python3 main.py [-h] -i inputfile -o outputfolder [-m modeltype] [-l length] [-c cutoff]
    
    required arguments
    
     -i, --input  : input fasta file.Please put the file in input folder or enter the full path
     -o, --output : output folder where predictions will be saved
    
    optional arguments
     
     -h, --help   : show this help message and exit
     -m, --model  : Machine-learning model to use for making predictions. Use 'general' for general predictions(default), 'viral' for viral prediction and 'bacterial' for bacterial predictions.
     -l, --length : Length of the peptides to be calculated from the sequence. Should be an integer >=4. For best results use peptide length = 20(default).
     -c, --cutoff : cut-off score (default = 0.5) for a peptide to be considered an epitope. Should be a number between 0 and 1.
  

• If you encounter a compatibility issue in sequencelist_representation.py related to KeyedVectors, you can fix it by going to /your_path/python3.7/site-packages/gensim/models/ldamodel.py. On line 57, replace from scipy.misc with from scipy.special

Input

The program takes a file with protein sequences as input. The input file should be in FASTA file format and should only use the 20 aa codes. The fasta file can contain a single or multiple protein sequences. If any sequence contains an invalid aa code or the length of the sequence is <4, then that sequence will be skipped while making predictions. Also, if the length of the sequence is < specified peptide length -l length, then the prediction will be made on the entire sequence.
Please put the input file in the input folder or enter the full path of the inputfile. Please take a look in the input folder to see an example of input fasta file: example.fasta

Output

The output files are created in the -o output folder. If the specified folder doesn't exist in the output path provided, it will be created. For each sequence in the fasta file, files with the name of the sequence and the suffix .epitopes.txt will be created in the output folder. The first column is the name of the sequence, second is the peptide, third is the starting postion of the peptide in the sequence, fourth is the ending postion of the peptide in the sequence, and the last column is the predicted probability score for the likelihood of that peptide being an epitope. By default, a predicted score of >= 0.5 is considered to be an epitope. To specify a different score cutoff, please use the -c cutoff argument. If the fasta file contains a single sequence then the predictions are also displayed on the commandline.

License

GPL-3.0 License

Citation

If you use EpitopeVec for your work, please cite as:

Bahai, A., Asgari, E., Mofrad, M. R. K., Kloetgen, A., & McHardy, A. C. (2021). EpitopeVec: linear epitope prediction using deep protein sequence embeddings. Bioinformatics, 37(23). https://doi.org/10.1093/bioinformatics/btab467

Contact

• Open an issue in the repository.
• Send an email to Akash Bahai (akashbahai@gmail.com).