Statistical Modelling of TCR Repertoires: Novel methods for analysis and guidance of immunotherapy

Research output: Book/ReportPh.D. thesis

Abstract

The research presented in this thesis involves development of novel computational tools for understanding and modelling the antigen-specific T cell adaptive immune response.
Adaptive immunity has a crucial role in immune protection as it is able to target and eliminate threats without incurring damage on the surrounding healthy tissue. In particular, T cells are able to engage foreign or malformed self antigens through specific recognition of the T cell receptor (TCR) expressed on their surface. Molecular pathways involving major histocompatability complex (MHC) expression ensure that a broad pool of antigen peptide fragments, epitopes, are displayed for the T cells to survey and possibly react to. To assure wide-ranging recognition of all possible epitopes, TCR undergo a stochastic process of generation and subsequent selection in order to prevent autoimmunity and secure binding to the individuals MHC molecules. This process generates a diverse pool of TCRs that differ in their sequences and in numbers between individuals. Due to the high complexity of the relationship between T cells and their targets, it is difficult to compare adaptive immune responses between individuals in a therapeutic setting.
The work presented in this doctoral thesis dives into dissecting the structure of T cell receptor responses on a global and a local scale. The assessment of global T cell responses have a principal importance in predicting disease outcomes. A widely used approach is to estimate TCR diversity in response to treatment or immunisation. Currently used diversity measures monitor clonal expansions of T cells as an indication of antigenic engagement. It has been demonstrated that T cells responding to the same antigen share sequence similarity. However, T cell repertoire diversity estimates lack information on TCR similarity, which is an important aspect of antigen recognition. The computational method presented in this thesis, TCRDivER, simultaneously incorporates both facets of T cell responses: clonality and similarity. By taking similarity into account, TCRDivER addresses some of the main drawbacks in repertoire comparison and allows for a view into the global repertoire structure. We show that TCRDivER, as a comprehensive method of surveying repertoire structure reveals unique features of TCR repertoires allowing classification between healthy repertoires and repertoires following immunisation.
One challenge of repertoire classification is the use of existing methods for comparing T cell similarity and prediction of T cell targets. Most similarity metrics rely on methods developed for evolutionary related proteins. T cell binding surfaces are diverse in sequence due to the stochastic nature of T cell development and the wide range of possible antigens they need to engage. Therefore, conventional techniques of estimating protein similarity and structure are less than optimal for these diverse proteins. The second model presented in this thesis, Tdist, aims to tackle this problem through the use of a machine learning algorithm. Tdist combines elements of TCR sequence and structural similarity in an attempt of extracting more information from a range of conventional similarity measures. This work hinges on the recent experimental advances that provide information on both T cell and their epitope target sequences. The advent of single cell sequencing promises even more resolution in T cell similarity metrics. However, due to the scarcity of such data Tdist is aimed at predicting T cell similarities from the widely available _ chain TCR sequences. Even though the work presented is a part of an ongoing study, it shows promise in linking T cell sequence similarity with epitope prediction.
As a whole, the work presented in this thesis aims to uncover inner-workings of T cell adaptive immune response in the mentioned global and local resolution. It paves the way for future work in developing personalised medicine approaches by illuminating some of the relationships guiding T cell immune responses.
Original languageEnglish
PublisherDTU Health Technology
Number of pages188
Publication statusPublished - 2020

Fingerprint

Dive into the research topics of 'Statistical Modelling of TCR Repertoires: Novel methods for analysis and guidance of immunotherapy'. Together they form a unique fingerprint.

Cite this