Using antibody next generation sequencing data to aid antibody engineering

Thesis

Future successful exploitation of antibodies as diagnostic and therapeutic agents will greatly benefit from an increased understanding of natural B-cell receptor (BCR) repertoire diversities. The advent of next-generation sequencing of immunoglobulin genes (Ig-seq) has made it possible to sequence large snapshots of BCR repertoires in a single experiment.

In the results chapters of this thesis, we begin by describing a method (AntiBOdy Sequence Selector, “ABOSS”) for filtering BCR repertoire data, which considers the structural viability of each sequence and is orthogonal to all other current methods (Chapter 2). ABOSS leverages the presence/absence of a conserved disulphide bridge found in antibodies as a way of both identifying structurally viable BCR sequences and estimating the sequencing error rate. We show that this method is able to identify structurally impossible sequences missed by common error-correction methods.

Next, we describe the development of Observed Antibody Space (OAS), the first resource that curates BCR sequences from publicly available studies. As of October 2020, OAS contains more than 1.9 billion sequences from 85 studies. In OAS, all BCR repertoire sequences are annotated and profiled for structural viability.

We next describe the development of a novel method (SAAB+) to interrogate complete BCR repertoires at the structural level (Chapter 4). SAAB+ annotates large portions of BCR repertoires with three-dimensional information by mapping sequences to crystallographically solved antibody structures. By applying SAAB+ to BCR repertoires in OAS we, for the first time, document repertoire structural changes along the B-cell maturation axis in humans and mice.

In the final experimental chapter, we describe our work in COVID-19 research where we have compared the structural and sequence diversities of SARS-CoV-2 BCR repertoires to healthy repertoires deposited in OAS. We also outline the development of the first organised database (CoV-AbDab) that curates all publicly available anti-SARS-CoV-2 antibodies in a standardised format.

Finally, we discuss how recent developments in paired-chain Ig-seq platforms and deep learning algorithms could have a lasting impact on established Ig-seq analysis pipelines. We also outline how the tools described in this thesis can be combined with these field-disruptive technologies to advance our understanding of the immune system and improve computational antibody engineering.

Authors: Kovaltsuk, A

• DOI: 10.5287/ora-2nn77mnjy
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: NGS; BCR
• Subjects: immunoinformatics