EPIC-TRACE: predicting TCR binding to unseen epitopes using attention and contextualized embeddings
Dani Korpela, Emmi Jokinen, Alexandru Dumitrescu, Jani Huuhtanen, Satu Mustjoki, Harri Lähdesmäki · 2023
Read EPIC-TRACE: predicting TCR binding to unseen epitopes using attention and contextualized embeddings by Dani Korpela, Emmi Jokinen, Alexandru Dumitrescu, Jani Huuhtanen, Satu Mustjoki, Harri Lähdesmäki (2023) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-10963061.txt.
Description
Korpela and co-authors build a machine learning model predicting whether a T cell receptor will bind a given target, including targets the model has never seen. T cells need this binding step to start an immune attack, so predicting it matters for vaccines and cell therapies. The hard part is generalizing: most models fail on new peptide targets. Their system, EPIC-TRACE, reads both chains of the receptor plus the peptide and its presenting molecule, using protein language model embeddings and attention layers.
The authors show which inputs matter most and find adding targets with only a few known receptors helps training. On standard databases, the model compares well with the rivals; the code is freely available. For immunology teams screening candidate targets, the tool offers predictions reaching beyond the pairs already measured in the lab. That reach is what therapy design needs.
Catalog record
| CREATOR | Dani Korpela, Emmi Jokinen, Alexandru Dumitrescu, Jani Huuhtanen, Satu Mustjoki, Harri Lähdesmäki |
|---|---|
| YEAR | 2023 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |