Reference-based comparison of adaptive immune receptor repertoires

Abstract

B and T cell receptor (immune) repertoires can represent an individual's immune history. While current repertoire analysis methods aim to discriminate between health and disease states, they are typically based on only a limited number of parameters. Here, we introduce immuneREF: a quantitative multidimensional measure of adaptive immune repertoire (and transcriptome) similarity that allows interpretation of immune repertoire variation by relying on both repertoire features and cross-referencing of simulated and experimental datasets. To quantify immune repertoire similarity landscapes across health and disease, we applied immuneREF to >2,400 datasets from individuals with varying immune states (healthy, [autoimmune] disease, and infection). We discovered, in contrast to the current paradigm, that blood-derived immune repertoires of healthy and diseased individuals are highly similar for certain immune states, suggesting that repertoire changes to immune perturbations are less pronounced than previously thought. In conclusion, immuneREF enables the population-wide study of adaptive immune response similarity across immune states.

Keywords: computational immunology; diagnostics; disease; health; immune repertoire.

Graphical abstract

Figure 1

| Reference-based comparison of adaptive immune receptor repertoires (AIRRs) (A) The complexity of AIRRs spans the frequency, motif, and feature space to each, of which distinct repertoire features may be attributed: the immune information stored in AIRRs is multidimensional. A longstanding question in the AIRR field is how to quantitatively measure inter-sample (sample, e.g., individual, immune cell population) AIRR similarity by accounting for AIRR feature multidimensionality in the effort to understand the distribution of inter-sample AIRR similarity across different immune events or immune cell populations. (B) We set out to develop an AIRR similarity measure that is sensitive, captures maximal immune information, and is sufficiently flexible to allow future integration of additional repertoire features (extensibility). (C) Each AIRR is represented as a node in a similarity network. The edges connecting the nodes represent the similarity score between the AIRR based on the six repertoire features. The immuneREF approach establishes interpretability on different levels: (1) from a single-feature perspective, the application of spider plots allows for an interpretable comparative analysis between repertoires, enabling the user to interpret the result observed in the condensed network on a per feature basis. (2) From the condensed feature network perspective, a major novelty introduced by the immuneREF workflow is the ability to combine established repertoire features into a common coordinate system. This transformation allows the combination of trends across features into a single condensed network that represents pairwise-cross-feature similarities. These pairwise similarities allow for the identification of subsets of more similar or aberrant repertoires. Interpretability on both features means allowing comparison to other repertoires and to simulated ones (of which we know the repertoire structure as ground truth), thus creating similarity equivalence classes. Equivalence classes create sets of reference repertoires, which enable interpreting the repertoire structures of other repertoires solely based on the immuneREF similarity score.

Figure 2

immuneREF measures immune repertoire similarity with high sensitivity using features that capture immune repertoire biology We simulated 200 immune repertoires using 40 different parameter combinations (in quintuplicate). (A) Hierarchical clustering visualizes the sensitivity of immuneREF by the successful grouping of immune repertoires that were simulated with slightly different parameters (composite network; see main text for details). (B) Network visualization with simulated repertoires as nodes and weighted edges between repertoires of similarity values above the upper quartile. (C) Quantification of mutual information among immune repertoire features. (D) Change in mean similarity of composite networks of increasing number of features. t test significance values are defined as ns: p ≥ 0.05, ∗: p < 0.05, ∗∗: p ≤ 0.01, ∗∗∗: p ≤ 0.001, ∗∗∗∗: p ≤ 0.0001.

Figure 3

The similarity landscape of simulated repertoires defines reference repertoires (A) Baseline similarity between replicates for repertoires simulated using default immuneSIM parameters (see Table S1) is ≥0.96 for five of six features, with the convergence feature being the exception by definition at ≤0.09. Bar graphs show mean SEM across replicates. (B) Repertoire similarity distribution in a condensed network across the various evaluated parameter range. Across cohorts, similarity scores have a broad range, whereas within cohorts the range is more restricted. (C) Workflow to determine representative repertoires per cohort going from many-to-many to a one-to-one comparison. (D) Local similarity distribution per species/receptor combination enables situating each repertoire based on its connectivity with respect to neighbors in the same cohort. (E) Comparing repertoires with maximal local similarity in their cohort visualizes the commonalities between receptor types; here the Murine IgH repertoire with maximal local similarity serves as a reference repertoire. The plot visualizes the similarities of each non-reference repertoire to the Murine IgH reference.

Figure 4

Application of immuneREF to 1,522 experimental repertoires (A) Similarity landscape of experimental (human, TCR) repertoires across three immune states (healthy, 439 repertoires; rheumatoid arthritis, 206 repertoires; and systemic lupus erythematosus, 877 repertoires). (B) Network visualization of the 1,721 nodes and weighted edges between repertoires of similarity scores (at three cutoff levels, 25%, 50%, and 75% top edge weights). (C) Distribution of similarity scores across the entire network and per immune state shows different degrees of within-cohort homogeneity. (D) Distribution of local similarity values per repertoire, faceted by cohort. (E) Comparison of the repertoires with the highest local similarity per immune state and an immuneSIM reference repertoire (default immuneSIM parameters; see Table S1).