This research reconstructs viral transmission trees using genomic sequencing data to study how human behavior shapes infectious disease outbreaks. Analyzing COVID-19 transmission in Iceland revealed differences in infectiousness across quarantined and demographic groups, informing vaccine distribution strategies that improved population-level protection and influenced national public health policy.
This research applies large language models to decode and design proteins by treating amino acid sequences as biological languages. By identifying hidden structural and functional patterns across massive protein datasets, the work enables creation of novel proteins for medicine, cancer therapy, carbon capture, and environmental remediation beyond naturally evolved biological systems.
This research uses natural language processing techniques to uncover evolutionary relationships between ancient proteins. By analyzing contextual patterns among amino acids, the new computational tool can identify connections between proteins that diverged billions of years ago, helping scientists reconstruct the history of early microbial life and Earth’s biological evolution.
This research uses spatial transcriptomics to map interactions between T cells, cancer cells, and immunosuppressive cells in tumours. Findings suggest cancer suppresses immune responses by surrounding and weakening T cells. The work aims to improve immunotherapy and enable personalised cancer treatment through detailed tumour mapping.
This thesis examines how octopuses respond to climate change at a molecular level, focusing on ocean acidification and RNA editing. Rising temperatures harm octopus reproduction, growth, and survival, while acidification produces mixed effects—some species show stress, yet others demonstrate resilience. Cephalopods overall appear more tolerant of acidification than fish, raising questions about the mechanisms behind this adaptability. Thousands of acidification-responsive edits disproportionately affect C2H2 zinc finger regulators, altering predicted binding targets, including nuclear pore components implicated in stress responses.
This research applies machine learning to genetic data to distinguish harmless DNA variations from cancer-causing mutations. By treating DNA like a crime scene, the model learns to identify which genetic changes truly drive breast cancer risk, supporting more accurate diagnosis and informed clinical decision-making.
This research investigates HMGN proteins, which organize the genome and help cells access the correct genes. By mapping their activity and removing them with CRISPR, the study shows that HMGNs act as DNA “librarians.” Their dysfunction leads to gene misregulation linked to many diseases.
About 8% of the human genome originates from ancient viruses. This research uses bioinformatics and evolutionary comparisons to understand why viral DNA persists and how cells silence it through DNA methylation. Identifying how genomes separate useful from non-functional DNA helps clarify which genetic elements matter for human health and disease.
This research develops a computational method for detecting hidden RNA viruses within existing RNA sequencing datasets. By identifying conserved viral protein signatures, the approach enables large-scale discovery of previously unknown viruses, improving understanding of viral diversity, disease mechanisms, and future opportunities for diagnostics, surveillance, and antiviral treatment development.
Mashpit is a portable genome-search tool that runs on a Raspberry Pi, enabling rapid, offline screening of Salmonella genomes. Using MinHash sketches, it scans hundreds of thousands of genomes in seconds, offering small or low-resource labs a fast, accessible way to identify related isolates before performing high-resolution follow-up analyses.