Kieran D. Lamb, Joseph Hughes, Spyros Lytras, Francesca Young,
Orges Koci, James C. Herzig, Simon C. Lovell, Joe Grove, Ke Yuan, and David L. Robertson
Nature Communications, 17, Article 69569 • DOI:
10.1038/s41467-026-69569-9
We investigate how Protein Language Model (PLM)representations enable assessment of variation due to
mutation. Applied to the
SARS-CoV-2 spike protein via in silico deep mutational scanning (DMS), the PLM ESM-2 captures evolutionary
constraints directly from sequence context, recapitulating what normally requires MSA data. Unlike other
state-of-the-art methods which require protein structures or multiple sequences for training, we show what
can be accomplished using an unmodified pretrained PLM. Applied to SARS-CoV-2 variants across the
pandemic, we demonstrate that ESM-2 representations encode the evolutionary history between variants, as
well as the distinct nature of variants of concern upon their emergence, associated with shifts in
receptor binding and antigenicity. ESM-2 likelihoods can also identify epistatic interactions among sites
in the protein. Our results here affirm that PLMs like ESM-2 are broadly useful for variant-effect
prediction, including unobserved changes, and can be applied to understand novel viral pathogens with the
potential to be applied to any protein sequence, pathogen or otherwise.
@article{lamb2026single,
title={From single-sequences to evolutionary trajectories: protein language models capture the evolutionary potential of SARS-CoV-2},
author={Lamb, Kieran D and Hughes, Joseph and Lytras, Spyros and Young, Francesca and Koci, Orges and Herzig, James C and Lovell, Simon C and Grove, Joe and Yuan, Ke and Robertson, David L},
journal={Nature Communications},
volume={17},
pages={69569},
year={2026},
publisher={Nature Publishing Group},
doi={10.1038/s41467-026-69569-9}
}
Kieran D. Lamb, Martha M. Luka, Megan Saathoff, Richard J.
Orton, My V. T. Phan, Matthew Cotten, Ke Yuan, and David L. Robertson
PLOS Computational Biology, 20(1): e1011795 • DOI:
10.1371/journal.pcbi.1011795
We show that both public health measures and virus properties are associated with the rise and fall of
regional SARS-CoV-2 reported infection numbers with regional differences attributable to the extent of
vaccine usage and the effectiveness of public health measures. In our mutational signature analysis, using
non-negative matrix factorisation, we detected three distinct mutational signatures that can be putatively
attributed to the action of specific host antiviral molecules. Interestingly, we observe a shift in
mutational signature activity from predominantly Signature 1 changes to an increasingly high proportion of
changes consistent with Signature 2. These mutation patterns influence SARS-CoV-2âs evolutionary capacity,
the available genetic variation that selection can act on, and so can be linked to the mutations defining
the variants of concern responsible for the distinct SARS-CoV-2 infection waves. The dominant types of
nucleotide substitutions involved indicate that much of the mutation and hence variation come from the
action of the host immune response rather than replication errors since the virus has an error correction
system.
@article{lamb2024mutational,
title={Mutational signature dynamics indicate SARS-CoV-2's evolutionary capacity is driven by host antiviral molecules},
author={Lamb, Kieran D and Luka, Martha M and Saathoff, Megan and Orton, Richard J and Phan, My VT and Cotten, Matthew and Yuan, Ke and Robertson, David L},
journal={PLOS Computational Biology},
volume={20},
number={1},
pages={e1011795},
year={2024},
publisher={Public Library of Science},
doi={10.1371/journal.pcbi.1011795}
}
Spyros Lytras, Kieran D. Lamb, Jumpei Ito, Joe Grove, Ke Yuan,
Kei Sato, Joseph Hughes, and David L. Robertson
Journal of Virology, 99(2): e01601-24 • DOI:
10.1128/jvi.01601-24
We discuss how a state-of-the-art artificial intelligence approach called protein language models (pLMs)
can be used for effectively analyzing pathogen genomic data. We highlight examples of pLMs applied to
predicting viral properties and evolution and lay out a framework for integrating pLMs into genomic
surveillance pipelines.
@article{lytras2025pathogen,
title={Pathogen genomic surveillance and the AI revolution},
author={Lytras, Spyros and Lamb, Kieran D and Ito, Jumpei and Grove, Joe and Yuan, Ke and Sato, Kei and Hughes, Joseph and Robertson, David L},
journal={Journal of Virology},
volume={99},
number={2},
pages={e01601-24},
year={2025},
publisher={American Society for Microbiology},
doi={10.1128/jvi.01601-24}
}
Kieran D. Lamb • Supervised by Prof. David L. Robertson
& Dr. Ke Yuan
Doctor of Philosophy (PhD) • University of Glasgow & University of Edinburgh (2020
– 2024)
Doctoral research developing novel machine learning architectures, matrix factorisation methods, and
protein language embeddings to investigate mutational processes, epistatic fitness landscapes, and the
evolutionary dynamics of resistance in viral genomes.
@phdthesis{lamb2024thesis,
title={Applying machine learning models to genome data to understand the evolution of drug resistance from virus to cancer evolution},
author={Lamb, Kieran D},
year={2024},
school={University of Glasgow and University of Edinburgh}
}
