Glastonbury Group
Human Technopole
Fondazione Human Technopole · Milan
The Glastonbury Group applies and develops machine learning methods to understand the genetic basis of a range of complex diseases. With the advent of population scale biobanking, rich data modalities are being collected on hundreds of thousands of individuals. One such data type is imaging data, a high dimensional and information rich modality for human phenotyping.
Approach
The Glastonbury Group is interested in applying machine learning techniques to any aspect of human genetic discovery (Variant → Gene → Phenotype). One of the key focuses the Glastonbury Group has is to extract cellular and disease mechanism phenotypes from histopathology images. Whole Slide Imaging (WSI) of tissue sections are routinely used to diagnose a range of diseases, such as Inflammatory Bowel Disease (IBD), various oncologies as well as neurological diseases post-mortem. By using both supervised and unsupervised machine learning techniques, it is possible to extract highly precise cellular and tissue phenotypes from such images and use them to characterise the cause and consequence of an underlying disease. By coupling such image derived phenotypes with genetic information (e.g. genotyping, exome, whole genome-sequencing), we can start to unpick mechanistically how a genetic variant contributes to increasing disease risk and how such risk is shared across diseases and phenotypes.
Variant
Genotypes, exomes and whole genomes across biobank-scale cohorts.
Gene
Prioritisation that says which gene at a locus is actually doing the work.
Phenotype
Cell and tissue measurements learned from images, not read off a label.
Research
Representation
Unsupervised, self-supervised and supervised methods, developed for and applied to large-scale biomedical imaging datasets.
Unsupervised, self-supervised and supervised methods, developed for and applied to large-scale biomedical imaging datasets.
Phenotype
Phenotypes learned directly from clinical data and imaging, without waiting for someone to define the label first.
Phenotypes learned directly from clinical data and imaging, without waiting for someone to define the label first.
Association
Development of scalable methods and pipelines for rapid GWAS and gene prioritisation.
Development of scalable methods and pipelines for rapid GWAS and gene prioritisation.
People
Group Leader
Dr Craig A. Glastonbury holds a PhD in computational biology from King’s College London (2013–2017). Before joining Human Technopole as a Research Group leader (06/2022), Craig was a lead machine learning researcher at the drug discovery company BenevolentAI (2019–2022) and a Postdoctoral research fellow at the University of Oxford (2017–2019).
Postdoc
Postdoc
PhD student
PhD student
PhD student
PhD student
PhD student
Postgraduate intern
Retreats
Retreat location · 2025
· 2025
Retreat location · 2025
· 2025
Wine tasting, retreat 2025
Wine tasting, retreat 2025
Cooking class, retreat 2025
Cooking class, retreat 2025
Retreat · 2025
From our first group retreat.
Contact
For questions about our research or possible collaborations, write to Craig.
PhD · Open
The Glastonbury Group uses functional genomics and single-cell technologies to map genetic risk variants (GWAS/Exome) to specific cell types, target genes, and disease mechanisms. By combining digital pathology, spatial transcriptomics (e.g., Xenium, Multiome), and patient-derived GI organoids, this project aims to:
PhD · Open
The Glastonbury Group leverages machine learning, digital pathology, and spatial transcriptomics to bridge human genetics (GWAS/Exome) with cell-type-specific disease mechanisms and precision diagnostics. By integrating routine clinical imaging with high-resolution spatial assays, our research focuses on three core themes: