ABSTRACT
B and T cells expand clonally in response to pathogenic infection, and their descendants, which share the same receptor sequence, can persist for years, forming the basis of immunological memory. While most T cell receptor (TCR) sequences are seen very rarely, ‘public’ TCRs are present in many individuals.
Using a combination of high throughput immunosequencing, statistical association of particular TCRs with disease status, and machine learning, we identified of a set of public TCRs that discriminates cytomegalovirus (CMV) infection status with high accuracy. This pathogen-specific diagnostic tool uses a very general assay that relies only on a training cohort coupled with immunosequencing and sophisticated data analysis. Since all memory T cell responses are encoded in the common format of somatic TCR rearrangements, a key advantage of reading T cell memory to predict disease status is that this approach should apply to a wide variety of diseases. The underlying dataset is the largest collection of TCRs ever published, including ∼300 gigabases of sequencing data and ∼85 million unique TCRs across 640 HLA-typed individuals, which constitutes by far the largest such collection ever published. We expect these data to be a valuable public resource for researchers studying the TCR repertoire.
