Linking immune receptor sequence to transcriptional states with clonotype neighbor graph analysis (CoNGA)

Abstract

Abstract Recent advances now allow for deep simultaneous profiling of T cell clonotypes, defined by T cell receptor (TCR) sequence, and phenotype, as reflected in gene expression (GEX) profile, surface protein expression, and epitope binding at the single-cell-level. However, there currently few tools available for unsupervised discovery of relationships between TCR sequence and cell phenotype. We hypothesized that by identifying correlations between “TCR neighborhoods”, defined by shared TCR sequence and GEX features, we could move beyond simply measuring GEX variation within clonal descendants and identify novel associations between T cell specificities and states. Previously, we introduced TCRdist, a measure for assessing inter-TCR similarity capable of identifying closely related clonotypes based on shared sequence features. Using TCRdist to quantify TCR similarity, we developed a graph-theoretic approach—clonotype neighbor-graph analysis or “CoNGA”—that identifies correlations between GEX profile and TCR sequence in an unbiased and automated manner through statistical analysis of GEX and TCR similarity graphs. Applying CoNGA, we uncovered novel associations between TCR and GEX space including a previously undescribed “natural lymphocyte” population of human blood CD8+ T cells; an association between TRBV gene usage and EPHB6 expression; and TCR sequence determinants of differentiation in thymocytes. These examples demonstrate that CoNGA is able to effectively deconvolve complex relationships between TCR sequence and cellular state. Conceptually, CoNGA could be extended to other clonally-related populations (e.g. B cells, tumors), and can easily incorporate other measurable features (e.g. ATAC-Seq).