Private Antibody Repertoires Are Public

Abstract

When faced with a given pathogen, the antibody response generally functions similarly across different people, 1-4 but the source of this similarity has been unclear. One hypothesis was that people share a high proportion of the same VDJ-recombined antibody genes, but this has been disproven. 5,6 An alternative is that people share a high proportion of functionally similar antibodies, 7,8 but testing this hypothesis requires a method for measuring functional similarity that scales to the millions of antibodies per repertoire and across multiple repertoires, which is impossible experimentally. We recently described a framework for doing so computationally, 9 which revealed that repertoires consist of loose overlapping functional classes of antibodies with similar antigen-binding capacities; 10-12 this framework allowed us to estimate a repertoire's antigen-binding capacity, , for the ideal target of any given antibody. Here, we show that this framework supports the second hypothesis, and provide the first comprehensive demonstration of overwhelming functional overlap between repertoires from 20 different individuals directly from sequence, without need of binding studies. Overlap is highest among the young and falls with age, due to the selective loss of antibodies that represent a core set of shared or "public" antigen-binding capacities. We reveal considerable heterogeneity in antigen-binding capacities for antibodies against influenza, HIV, and SARS-CoV-2, and show that while some of these classes shrink with age, others persist across individuals. These discoveries change our understanding of repertoire diversity and have implications for vaccine and therapeutic-antibody development, especially for the aged.Recent large-scale studies have shown that <1% of IgG heavy-chain CDR3s are measurably shared across people: the so-called public repertoire of shared sequences is small, dwarfed by the >99% of the repertoire that is "private" or effectively unique to a given individual. 5,6 The questions arise as to

Epitope profiling using computational structural modelling demonstrated on coronavirus-binding antibodies

“…Furthermore, given that individuals' naive repertoires typically share very few clonotypes [11,26] and yet are often found to respond to similar 'immunodominant' epitopes, it follows that multiple evolutionary routes may lead from low-moderate affinity naive BCRs to high affinity antibodies against the same antigen surface region. This is supported by statistical arguments showing the implausability of a purely "random repertoire" for an efficient immune response [25,27]. Epitope immunodominance could be rationalised via the existence of a more 'public' set of backbone structures in the BCR repertoire and the concept that BCRs with similar topologies and sufficient chemical complementarity engage the same epitope [24,25].…”

Repertoire-Scale Measures of Antigen Binding

“…Extraordinarily, for both IGH and TRB, the number of classes in the entire sampled population was not substantially larger than that in a single young, healthy individual. This result implies that the private repertoire is functionally public and that classes hold additional useful patterns, which future studies may reveal ( 18, 60–62 ).…”

ENRICH: Exploiting Image Similarity to Maximize Efficient Machine Learning in Medical Imaging

“…In addition, different algorithm choices as well as code optimizations can be explored to maximize the utility of ENRICH while minimizing time and computational load. Quantitative measures of similarity have been shown to add useful insights in other fields 37,38 . ENRICHment, in various forms, is expected to be a useful new avenue for decreasing labeling burden and speeding iterative training and testing of DL models in development.…”