Computational approaches to therapeutic nanobody design

Thesis

Nanobodies, or single-domain antibodies, have favourable characteristics that make them a promising novel therapeutic format. However, their unique structural and functional features also present distinct challenges for optimisation in therapeutic applications. Whilst computational approaches have significantly advanced the rational design of conventional antibodies, similar efforts for nanobodies remain limited. To fully realise the therapeutic potential of nanobodies, there is a pressing need for in silico methods tailored to their specific properties and associated challenges. Progress in this area is currently constrained by an incomplete understanding of nanobody behaviour in therapeutic contexts and the relatively limited availability of relevant data. In this thesis, we develop computational approaches to support the design of therapeutic nanobodies. We focus on characterising the binding and developability properties of nanobodies to establish a foundation for building bespoke tools that can identify, evaluate, and optimise novel candidates.

Firstly, we compare the binding sites of conventional antibodies and nanobodies to understand the structural basis of antigen recognition. By identifying residues most critical for binding, we provide a basis for multi-parameter optimisation, helping to distinguish positions that can be modified to improve developability without compromising specificity or affinity. To address the limited availability of nanobody data, we created PLAbDab-nano, a curated database of nanobody sequences and modelled structures sourced from patents and academic literature. We envisage that this resource can be used as a foundation for designing diverse nanobody libraries, for benchmarking and improving existing scaffolds by comparison to therapeutically relevant or optimised sequences, and for the development of computational models trained specifically on nanobody data. The PLAbDab-nano database was used to characterise developability traits in nanobodies, leading to the development of the Therapeutic Nanobody Profiler, a computational tool that evaluates developability using sequence- and structure-based features tailored to the unique properties of nanobodies. Collectively, this work aims to enable rational design and multiparameter optimisation of therapeutic nanobodies by combining structural analysis, improved data availability, and computational tools specific to this novel format.

Authors: Gordon, G

• DOI: 10.5287/ora-gajxzagak
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English