Speaker
Emanuele Loffredo
(Ecole Normale Superieure)
Description
T-cell receptor binding with the pMHC peptide is a fundamental step to activate the killing machinery of the adaptive immune system against pathogens. However, limited data -- due to the high specificity of T-cell receptors -- make the task of binding prediction very challenging. Here we propose to use Large Language Models to sample new peptide-specific T-cell receptor sequences; by leveraging the transfer-learning ability of Large Language model, we are able to generate sequences compatible with natural ones and use them to improve the predictive power of deep classifiers.
Primary author
Emanuele Loffredo
(Ecole Normale Superieure)
