Computational analysis of the response of immune repertoires to viruses – RESP-REP

We will quantify repertoire-wide statistical changes of the repertoire composition following vaccination, and use this quantification to build a score predictive of vaccination status. We will find TCR and BCR sequences that expanded significantly in response to the vaccine, by assessing statistical significance relative to the null expectation of experimental variability inferred from biological replicates. Based on examples of expanded sequences, we will then use machine-learning techniques to classify responding sequences from non-responding ones. We will validate our predictions using functional and binding assays. Lastly, we will attempt to identify T-cell receptors that are specific to the vaccine from a single sample (no longitudinal data) of the repertoire at the peak of the response, using measures of local sequence concentration relative to a background stochastic model of sequence generation. Finally we will apply this single-sample method to the discovery of TCR specificity to a wide range of auto-immune disesases from samples from the Klatzmann lab.

RESP-REP will yield new datasets of BCR and TCR sequences with known specificities to the yellow fever and influenza vaccines, respectively. It will also provide a procedure to score any yet unobserved sequences to determine its vaccine specificity. A broader outcome of the project will be pipelines delivered in software packages that will automatically detect specific TCR and BCR sequences from longitudinal data or from single samples taken shortly after an immune response. This software can then be used by other teams to characterize the response to other vaccines, infections, or auto-immune conditions.

Our computational tools will make it possible to systematically associate a known immune function to vast numbers of TCR and BCR, and could be brought to the clinic, where repertoire sequencing is becoming increasingly common, to automate and accelerate discoveries of these associations in a high- throughput way.

1. G. Isacchini, A.M Walczak, T. Mora, Armita Nourmohammad Deep generative selection models of T and B cell receptor repertoires with soNNia Proc Natl Acad Sci 118(14) e2023141118 (2021)
2. J. Marchi, M. Lässig, A.M. Walczak*, T. Mora* Antigenic waves of virus-immune co-evolution Proc Natl Acad Sci 118(27) e2103398118 (2021)
3. A.A. Minervina, E.A. Komech, A. Titov, M. Bensouda Koraichi, E. Rosati, I.Z. Mamedov, A.Franke, G.A. Efimov, D.M. Chudakov, T.Mora, A.M. Walczak, Y.B. Lebedev, M.V. Pogorelyy
Longitudinal high-throughput TCR repertoire profiling reveals the dynamics of T cell memory formation after mild COVID-19 infection eLife 10 e63502 (2020)

RESP-REP aims to turn the power of high-throughput immune repertoire sequencing into useful medical diagnosis tools, by predicting the immune function and specificity of lymphocyte receptors sequences. The immune system relies on its diverse repertoire of specific B- and T-cell receptors (BCR and TCR) to recognize the large diversity of pathogen-derived antigens. The composition of the BCR and TCR repertoires carries the footprint of past immune responses and is informative about the ability to fight future infections. The recent years have seen rapid technological progress in the deep sequencing of repertoires in single individuals, with the promise to read off the immune status of patients and predict their ability to respond to infections, vaccines, or their sensitivity to auto-immune diseases. Such diagnosis tools could inform more efficient vaccine design, new personalized medicine strategies, or better targeted cancer immunotherapy. However, a major roadblock hindering progress is our inability to predict which of the billions of billions of B or T cell receptors expressed in the body can recognize a given antigen.
To address this issue, in this project we propose to develop advanced computational techniques to predict specific TCR and BCR sequences from the analysis of longitudinal repertoire samples from human donors in response to vaccination, and possibly from single samples taken at the peak of the immune response. We will analyze repertoire datasets (already in our possession) of TCR in response to the Yellow-Fever vaccine, and of BCR in response to the influenza vaccine, sampled at several time points before and after vaccination. We will make computational predictions which we will test using functional assays.
Specifically, we will quantify repertoire-wide statistical changes of the repertoire composition following vaccination, and use this quantification to build a score predictive of vaccination status. We will find TCR and BCR sequences that expanded significantly in response to the vaccine, by assessing statistical significance relative to the null expectation of experimental variability inferred from biological replicates. Based on examples of expanded sequences, we will then use machine-learning techniques to classify responding sequences from non-responding ones. We will validate our predictions using functional and binding assays. Lastly, we will attempt to identify T-cell receptors that are specific to the vaccine from a single sample (no longitudinal data) of the repertoire at the peak of the response, using measures of local sequence concentration relative to a background stochastic model of sequence generation. RESP-REP will yield new datasets of BCR and TCR sequences with known specificities to the yellow fever and influenza vaccines, respectively. It will also provide a procedure to score any yet unobserved sequences to determine its vaccine specificity. A broader outcome of the project will be pipelines delivered in software packages that will automatically detect specific TCR and BCR sequences from longitudinal data or from single samples taken shortly after an immune response. This software can then be used by other teams to characterize the response to other vaccines, infections, or auto-immune conditions.
Our computational tools will make it possible to systematically associate a known immune function to vast numbers of TCR and BCR, and could be brought to the clinic, where repertoire sequencing is becoming increasingly common, to automate and accelerate discoveries of these associations in a high-throughput way.