Overview
Fungal disease kills an estimated 3.75 million people a year worldwide — six times more than malaria and three times more than tuberculosis. Yet despite this staggering toll, the genomes of even our most dangerous fungal pathogens remain frustratingly poorly understood.
About this opportunity
Aspergillus fumigatus, a World Health Organization critical-priority pathogen, causes invasive and chronic pulmonary aspergillosis with alarming mortality and few effective drugs. The bottleneck is not a lack of genome sequences — it’s that most genes still have no reliable annotation of what they actually do, particularly for those that are critical for viability and pathogenicity. This PhD sits at the frontier of a growing field: can autonomous, reasoning AI agents interpret and understand gene function by cross-checking sequence, structure, evolutionary conservation and experimental evidence, and then validate their conclusions against real biology?
This is a new Liverpool-Manchester collaboration, underpinned by two major Wellcome awards, which brings together AI-driven genomics (Liverpool, home to FungiDB, a primary fungal genomics database) with cutting-edge experimental fungal biology and functional genomics (Manchester). Crucially, the student will have access to a rare resource: a genome-wide barcoded gene-knockout library and pooled fitness screens, alongside clinical isolate phenotyping, genome sequences and metabolomics, generated within Manchester’s own Wellcome programme. This will enable any predictions made by AI to be checked against real experimental outcomes as they are generated.
The student will build three linked capabilities. First, a precise pan-genome map of A. fumigatus, pinpointing every gene across high-quality reference genomes and tracking its presence or absence across hundreds of genomes in public and local repositories — revealing which genes are core to the species and which are accessory, adaptable cargo. Second, a multi-evidence annotation engine that goes beyond standard bioinformatics tools, fusing protein structure prediction, ligand-binding pocket detection, pan-genome conservation patterns and functional screening data into a single confidence-scored picture of what each gene does. Third, an agentic AI layer, building on a working prototype already live in FungiDB, that reads across all this evidence, plus the scientific literature, to write rich, specific, evidence-backed statements of gene function, autonomously deciding what to check and where the evidence is weak.
The resulting tools will help researchers worldwide identify new antifungal drug targets by druggability, essential-gene status and selectivity over human proteins, while also uncovering the biology of how this fungus adapts, evades immunity, and causes acute versus chronic disease. The approach is designed to generalise well beyond one species, with a clear route into other fungal pathogens in FungiDB and parasite genomes across the wider VEuPathDB resource (malaria, cryptosporidiosis, leishmaniasis and more).
The student will be embedded in Liverpool’s VEuPathDB team, with structured AI/coding training and a large peer community (e.g. via our Computational Biology Facility), while also joining Manchester’s Bromley lab with an extended placement, giving genuine dual-institution training spanning agentic AI development and hands-on fungal experimental biology.