python(version=3.7.9); torch(version=1.8.0)；torchvision(version==0.2.2); numpy(version==1.18.5); pandas(version=1.2.4); tqdm(version=4.19.9); matplotlib(version=3.3.2); seaborn(version=0.11.1);weblogo(version=3.7.12)

It is recommended to use the linux system, here we used Centos Linux release 7.9.2009(Core) system.

We employed Tesla K80 and Intel(R) Xeon(R) CPU E5-2678 v3 @2.50GHz.

This is the 1th version which has been tested on independent antigen presentation and immunogenicity test datasets.

Download the source code and unzip it. Simply enter the directory of IEPAPI and install all required dependencies on your operating system:

cd ./IEPAPI-main
pip install -r dependencies.txt

• Organize your input information in the following format:

  

• Run the command to make a prediction：

  python IEPAPI_predict.py --input test.csv --output output_demo.csv

• --input: the first column is the sequence of the peptide, the second column is the name of the HLA, and the third column is the pseudo-sequence of the HLA. The file ". /data/pseudoSequence(ELIM).csv" records the pseudo sequences of all HLA isoforms used in this study. For other HLA isoforms, their pseudo sequences can be found in the file ". /data/NetMHCpan4.1/MHC_pseudo.dat".

• --output: the path where you want to save the prediction results. Here, the expected output will be saved in the file "./output_demo.csv".

  

• Run the command to obtain the immune epitope motif for the specified HLA：

  python IEPAPI_motif.py --MHC HLA-A*11:01 --MHCseq YYAMYQENVAQTDVDTLYIIYRDYTWAAQAYRWY --require_pdf True

• --MHC: the name of the MHC molecule

• --MHCseq: the pseudo-sequence of the MHC molecule

• --require_pdf: If set to False, only images in JPEG format are output. If set to True, images in PDF format will also be output.

• For HLA-A*11:01, both motifs for antigen presentation and immunogenicity will be generated in the format of heatmaps and sequence logos, which can reflect the immune epitope pattern.

  

a) Download all data from the Mendeley Data website (https://data.mendeley.com/datasets/fwxg5mgntn) and place the data in the ". /data/processed" directory

b）Training the IEPAPI-EL model

 python  main_train_Model_EL.py   --fold 0   --index 0
 python  main_train_Model_EL.py   --fold 1   --index 0
 python  main_train_Model_EL.py   --fold 2   --index 0
 python  main_train_Model_EL.py   --fold 3   --index 0
 python  main_train_Model_EL.py   --fold 4   --index 0

c）Training the IEPAPI-IM model

 python  main_train_Model_IM.py   --fold 0   --index 0
 python  main_train_Model_IM.py   --fold 1   --index 0
 python  main_train_Model_IM.py   --fold 2   --index 0
 python  main_train_Model_IM.py   --fold 3   --index 0
 python  main_train_Model_IM.py   --fold 4   --index 0

d）Make predictions for the test datasets

 python IEPAPI_predict.py --input ./data/processed/DataS3.csv --output ./output/results/DataS3_by_IEPAPI.csv
 python IEPAPI_predict.py --input ./data/processed/DataS4.csv --output ./output/results/DataS4_by_IEPAPI.csv
 python IEPAPI_predict.py --input ./data/processed/DataS5.csv --output ./output/results/DataS5_by_IEPAPI.csv