Mapping Deep Phenotypes in Atopic Dermatitis: A Machine Learning Approach to Understanding Disease Diversity
Background
Atopic dermatitis (AD) is the most common chronic inflammatory skin disease, affecting up to 20% of children worldwide. It is characterised by itchy and visible skin lesions and has a profound impact on sleep, psychosocial wellbeing, and quality of life. However, it has a highly variable clinical presentation which differs with age, ethnicity, and geography. Most existing studies have been conducted in predominantly white cohorts from high-income settings, leaving significant gaps in our understanding of how immune, environmental, and microbial factors interact to underlie the global disease diversity.
Novelty & Importance
The project uses a novel, data-driven approach to uncover hidden AD subtypes by integrating microbiome, metabolome, immune, clinical, and patient-reported data from two ethnically diverse cohorts from the UK and South Africa. Using unsupervised machine learning and integrative multi-omics approaches, we will identify distinct biological and clinical signatures underlying disease heterogeneity. The results will be displayed through an interactive Deep AD Phenotype Atlas, enabling clinicians and researchers to explore phenotype clusters and associated immune-microbiome biological signatures.
Aims & Objectives
1. To identify deep phenotypes in atopic dermatitis by integrating microbiome, clinical, patient-reported, and environmental data from two diverse cohorts.
2. To develop an interactive, user-friendly dashboard (Deep AD Phenotype Atlas) that visualises latent subgroups across multimodal data.
By bridging data science and clinical dermatology, this project will advance understanding of the biological mechanisms of AD, and contribute towards more equitable, personalised approaches to diagnosis and treatment of this commonest of all skin diseases.