mnielLab / NetTCR-2.0

NetTCR-2.0. Sequence-based prediction of peptide-TCR binding
33 stars 10 forks source link

NetTCR-2.0 enables accurate prediction of TCR-peptide binding by using paired TCRα and β sequence data

This repository contains the code and the data to train NetTCR-2.0 model.

License

NetTCR-2.0 is developed by Morten Nielsen's group at the Technical University of Denmark (DTU). NetTCR-2.0 code and data can be used freely by academic groups for non-commercial purposes. If you plan to use NetTCR or any data provided with the script in any for-profit application, you are required to obtain a separate license (contact Morten Nielsen, morni@dtu.dk).

For scientific questions, please contact Morten Nielsen (mniel@dtu.dk).

Data

This data folder contains the data files used to train NetTCR-2.0.

File description:

Train networks

You can train the NetTCR_ab models running

python nettcr.py --trainfile test/sample_train.csv --testfile test/sample_test.csv

This will print the predictions on the standard output or on a file (that can be specified with the option --outfile).

Both training and test set should be a comma-separated CSV files. The files should have the following columns (with headers): CDR3a, CDR3b, peptide, binder (the binder coulmn is not required in the test file). See test/sample_train.csv and test/sample_test.csv as an example.

NetTCR server

NetTCR-2.0 is also availavble as a web server at https://services.healthtech.dtu.dk/service.php?NetTCR-2.0. The server offers the possibility to evaluate pre-trained models on new data. See Instructions tab for more information.

Citation

Montemurro, A., Schuster, V., Povlsen, H.R. et al. NetTCR-2.0 enables accurate prediction of TCR-peptide binding by using paired TCRα and β sequence data. Commun Biol 4, 1060 (2021). https://doi.org/10.1038/s42003-021-02610-3