Abstract Mucosal-associated invariant T (MAIT) cells are a group of unconventional T cells that mainly recognize bacterial vitamin B metabolites presented on MHC-related protein 1 (MR1). MAIT cells have been shown to play an important role in controlling bacterial infection and in responding to viral infections. Furthermore, MAIT cells have been implicated in different chronic inflammatory diseases such as inflammatory bowel disease and multiple sclerosis. Despite their involvement in different physiological and pathological processes, a deeper understanding of MAIT cells is still lacking. Arguably, this can be attributed to the difficulty of quantifying and measuring MAIT cells in different biological samples which is commonly done using flow cytometry-based methods and single-cell-based RNA sequencing techniques. These methods mostly require fresh samples which are difficult to obtain, especially from tissues, have low to medium throughput, and are costly and labor-intensive. To address these limitations, we developed sequence-to-MAIT ( Seq2MAIT ) which is a transformer-based deep neural network capable of identifying MAIT cells in bulk TCR-sequencing datasets, enabling the quantification of MAIT cells from any biological materials where human DNA is available. Benchmarking Seq2MAIT across different test datasets showed an average area-under-the-receiver-operator-curve (AU[ROC]) &gt;0.80. In conclusion, Seq2MAIT is a novel, economical, and scalable method for identifying and quantifying MAIT cells in virtually any biological sample.

Seq2MAIT: A Novel Deep Learning Framework for Identifying Mucosal Associated Invariant T (MAIT) Cells