Developing predictive models of treatment responsiveness in ulcerative colitis from spatial transcriptomics and electronic health records

Ulcerative colitis (UC) affects nearly one million adults in the United States. Although many effective medications are available, selecting the right treatment for an individual patient is still largely trial and error. We aim to enable precise, personalized UC treatment by identifying biological and clinical signals that predict therapeutic response at the outset. Building on our prior work identifying predictors of response to vedolizumab, we will extend our approach to the most commonly used UC therapies, including anti-TNF agents, IL-23 inhibitors and JAK inhibitors. Our multidisciplinary team will use our combined expertise in immunology, clinical care and artificial intelligence (AI) to study archived colon biopsy samples from UC patients with known treatment outcomes. We will measure gene activity and immune cell organization in standard archival material from patient tissue and apply machine learning and AI methods to identify predictive patterns of treatment response. We ultimately aim to translate these findings into clinically available measures in the electronic health record (EHR) and simplified laboratory tests. By developing validated predictive models, this work has the potential to reduce trial-and-error treatment burden, helping patients receive the right treatment earlier.