NeoMHCI

NeoMHCI is a deep learning tool designed for the prediction of multi-allele and single-allele MHC-I ligand presentation as well as neoepitope immunogenicity.

All data utilized in this research is available for free download on our Mendeley repository.

Requirements

The following packages and their versions are required:

• python == 3.8.8
• pytorch == 1.7.1
• numpy == 1.19.2
• click == 7.1.2
• ruamel.yaml == 0.16.12
• tqdm == 4.56.0
• fair-esm == 0.4.2

Usage

The python script main.py can execute two types of task modes: test and motif.

Test

Example command:

python main.py -d ./config/data.yaml -m ./config/model.yaml --mode test -s 0 -n 10 -g 0

In data.yaml, the following data definitions exist:

• mhc_seq: Defines the pseudo sequence for each MHC-I molecule.
• cell_line: Defines the real MHC-I molecules corresponding to multi-allele or single-alele names.
• test: Data to be tested.
• rawdata: Defines the hundred thousand random peptides used for drawing motifs.

When testing new single allele data or multi-allele data, ensure the data format is consistent with that in ./data/example_data.txt. Additionally, ensure the new MHC-I molecule or combinations are listed in the file corresponding to cell_line (default ./data/allelelist), and that the 34-mer pseudo sequence for the new MHC-I molecule is provided in the file corresponding to mhc_seq (default ./data/MHC_pseudo.dat).

The output results will be stored in ./outputs.

Motif

Example command:

python main.py -d ./config/data.yaml -m ./config/model.yaml --mode motif -s 0 -n 10 -g 0 --allele CPH-08-TISSUE --motif_len 9

You can specify the single allele or multi-allele to output the top-1% (1000 peptides) by setting the --allele option. The --motif_len option defines the length of the chosen motif, with a maximum length of 15.

The output results will be placed in ./motifs.

Declaration

This tool is free for non-commercial use. For commercial use, please contact Mr. Wei Qu and Prof. Shanfeng Zhu (zhusf@fudan.edu.cn).