This proposal extends our 2022 UK Biobank (UKB) project (ID: 6508), on target discovery for age-related diseases. Since then, we have developed Rosalind, a deep learning model to predict gene regulatory effects from sequence, building on Enformer (Avsec et al., 2021 ; https://doi.org/10.1038/s41592-021-01252-x). We applied it to UKB whole-genome sequencing (WGS) data and integrated results with multi-omics and clinical data from our in-house study (https://www.osteomics.co.uk/). Functional target validation is underway.
We now aim to develop a unified strategy for target discovery across chronic diseases, including musculoskeletal (e.g., osteoarthritis, osteoporosis), fibrotic (e.g., systemic sclerosis) and metabolic (e.g., obesity) conditions. These share mechanisms such as inflammation, tissue remodeling, and senescence.
Building on evidence that genetically-supported targets are more likely to succeed (Minikel et al., 2024; https://doi.org/10.1038/s41586-024-07316-0), we will address two questions: (i) which variants drive coding or regulatory changes linked to disease, and (ii) how these insights inform mechanisms of action and target prioritization.
We will refine Rosalind using in-house genomic data – an approach shown to uncover regulatory mechanisms and disease risk (Zhou et al., 2024; https://pubmed.ncbi.nlm.nih.gov/39463940/), then apply it to UKB WGS cohort to assess variant functional impact, and integrate these with phenotypes (e.g. DEXA-derived BMD), proteomics, and disease endpoints. We will also study carriers of predicted loss-of-function or regulatory-impact variants to assess therapeutic potential, and safety through phenome-wide associations.
UKB’s paired genetic, molecular and clinical data are essential to our integrative approach. Specifically, we request access to genetic data (WGS, WES, genotyping) to capture relevant variants, and rich phenotypic data (imaging, proteomics biomarkers), to enable mechanistic insight.