Summary
Deep learning methods fail to identify new epitopes for T-cell receptors. They only generalize to very similar epitopes found in the training set (1). This could be applicable to B-cell receptors and antibodies as well.
See also
- DL models excel at finding pockets but not docking into them
- DL structure prediction methods cannot predict CDR conformations unseen during training
1.
Croce G, Bobisse S, Moreno DL, Schmidt J, Guillame P, Harari A, et al. Deep learning predictions of TCR-epitope interactions reveal epitope-specific chains in dual alpha T cells. Nature Communications. 2024;15(1). Available from: https://doi.org/10.1038/s41467-024-47461-8