Back

Machine learning model interpretations explain T cell receptor binding

Carter, B.; Krog, J.; Birnbaum, M. E.; Gifford, D. K.

2023-08-16 immunology
10.1101/2023.08.15.553228 bioRxiv
Show abstract

T cells mediate immune responses against pathogens and cancer through T cell receptors (TCRs) that recognize foreign peptides presented on the cell surface by Major Histocompatibility Complex (MHC) proteins. TCRs carry enormous diversity and differ across individuals, and mechanisms that determine TCR-pMHC binding are poorly understood. The ability to predict TCR-pMHC interactions would accelerate development of cellular therapeutics to design TCRs that specifically bind to a target of interest. We designed a randomized library of 196 TCR CDR3{beta} sequences and experimentally evaluated their affinities for the Tax-A02 peptide-MHC target. We trained ML models that predict TCR binding to Tax-A02 from TCR sequence and used model interpretation to identify TCR sequence features associated with binding. We found these features accurately mirror the true sequence features in our experimental data.

Matching journals

The top 2 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.