Uncertainty-aware quantitative analysis of the structure and dynamics of T cell receptor repertoires

Abstract

Diversity and dynamics of immune cell receptor repertoires (IRRs) are two factors at the functional heart of adaptive immunity that together make IRRs difficult to grasp. Moreover, measurements are compounded by various sources of experimental noise. Here we propose a computational framework (ClustIRR) for uncertainty-aware quantitative analysis of IRR structure and dynamics. ClustIRR maps multiple IRRs across replicates, time points, or conditions onto a joint graph induced by immune receptor sequence similarity. It then detects communities on the joint graph (CJs). Based on CJs as reference structures across IRRs, ClustIRR then performs quantitative Bayesian analyses of differential CJ occupancy. Additionally, ClustIRR integrates single-cell gene expression data to link community expansion with transcriptional activation signatures. We demonstrate the capabilities of ClustIRR with the joint analysis of multiple T cell receptor repertoires in several example applications: (1) quantitative changes due to antigen challenge, (2) longitudinal dynamics during cancer immunotherapy, (3) V(D)J recombination biases in human vs murine repertoires that pre-adapt IRRs for pathogen responses. ClustIRR is freely available as open source software from the bioconductor repository.