TCREMP: A Bioinformatic Pipeline for Efficient Embedding of T-cell Receptor Sequences from Immune Repertoire and Single-cell Sequencing Data.

T-cell receptors (TCRs) expressed by T-lymphocytes can recognize antigens presented on the surface of our cells to tell normal cells from infected and malignant ones. An extremely diverse set of TCRs is generated during T-lymphocyte formation and can be surveyed using high-throughput sequencing to tell the story of past and ongoing immune responses. Proper handling of TCR sequencing data and using it to trace antigen specificities is a challenging bioinformatic task. Here we present TCREMP (TCR embedding via Prototypes), a pipeline to digitize TCR amino acid sequences based on their similarity to commonly observed sequences that can provide a reference point for comparative analysis and aid in decoding antigen specificity prediction across TCR repertoire samples that one may find useful for adaptive immunity studies. Our results show that TCREMP embeddings can efficiently distinguish distinct antigen-specific T-cell populations based on single-cell and Major Histocompatibility Complex (MHCs)-multimer-sorted repertoire sequencing data. TCREMP is freely available at https://github.com/antigenomics/tcremp.