The TCR Motif Atlas: a database of specificities for TCRs recognizing viral, cancer, and autoimmune epitopes

T-cell recognition is elicited by the binding between T-cell receptors (TCRs) and epitopes. Many machine learning approaches have been proposed to model TCR-epitope recognition specificity, but most of them act as black boxes. To visualize TCR-epitope recognition specificity, we developed the TCR Motif Atlas (https://tcrmotifatlas.unil.ch). This database builds upon the probabilistic framework of TCR specificity profiles (TSPs) to visualize V and J genes as well as complementarity-determining Region 3 (CDR3) lengths and amino acids enriched in epitope-specific TCRs with respect to baseline TCR repertoires for hundreds of epitopes. An interactive interface enables users to build TSPs based on their own set of TCR sequences to investigate epitope recognition specificity or to explore potential biases in TCR repertoires of known or undetermined specificity. The TCR Motif Atlas further provides links to curated compilations of epitope-specific TCRs and crystal structures of TCR-epitope complexes.