Computational Methods to Design Broad-Spectrum Medical Countermeasures Against Antigenically Diverse Pathogens

Infectious diseases caused by rapidly mutating pathogens, such as coronaviruses and influenza viruses, pose substantial challenges for global health. Mutations in these pathogens can change their antigens—which are molecules recognised by T-cells and antibodies of the immune system. This antigenic diversity can reduce the effectiveness of existing vaccines and treatments. This dissertation explores how computational approaches, such as graph-based and deep learning methods, can be used to design broad-spectrum vaccines and antibodies that are effective against a wide range of pathogen variants. I focus on two complementary projects: 1) designing single-antigen vaccines to induce broad-spectrum T-cell responses, and 2) developing a deep learning model to predict antibody-antigen binding. First, I introduce Spectravax, a computational framework to design broad-spectrum vaccines optimised to account for genetic diversity in both pathogen and host populations. I applied Spectravax to seven coronaviruses and influenza A virus target antigens and demonstrated that Spectravax-designed antigens are predicted to have improved coverage of host and pathogen populations compared to existing wild types and computational designs. Experimental validation in mice confirmed these predictions, with the Spectravax nucleocapsid antigen being the first computationally designed antigen shown to elicit immune responses to SARS-CoV-1, SARS-CoV-2, and MERS-CoV—the three coronaviruses responsible for major outbreaks in humans since 2002. Second, I trained and evaluated a deep learning method to predict antibody-antigen binding affinity solely from their protein sequences. I assessed this model’s performance on related tasks, such as SARS-CoV-2 neutralisation prediction and tested its sensitivity to point mutations. I offer these methods to assess affinity predictors, which are crucial for the design of broadly neutralising antibodies. Overall, this work demonstrates that computational methods can be used for the rational design of broad-spectrum medical countermeasures. Such advancements pave the way for the development of next-generation vaccines and antibodies to improve global health security by reducing the burden of infectious diseases and mitigating the risk of pandemics.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Palmer, Phil

Advisor dc:contributor.advisor

• Heeney, Jonathan

• Bioinformatics
• Broad-spectrum vaccines
• Computational immunology
• Coronavirus
• Deep learning
• Vaccine design

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/7a00c742-3e4a-4345-a9da-ade8035a2284/download
• https://creativecommons.org/licenses/by/4.0/

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.119283

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/385816