Development and validation of machine learning optimized predictive models for response to different biologic agents in patients with Crohn's disease and ulcerative colitis

Project: Research project

Project Details


has created a novel methodological approach for building predictive models for biologics in IBD. The predictive models generated using this methodological approach for vedolizumab (anti-integrin biologic) in Crohn’s disease (CD) and ulcerative colitis (UC) accurately and very significantly (p<0.001), stratify inter-individual differences in: biologic drug concentrations (biologic exposure), clinical response and onset of action (efficacy), response to escalating biologic dosing (optimization), and progression to surgery while on active therapy (complication).2 Prior literature has demonstrated that core sets of predictors consistently emerge across biologic studies in IBD, however, the relative predictive importance or impact seems to vary across biologics and is not well categorized.3, 4 The candidates overarching hypothesis for this proposal is that his methodological approach for model building, when applied to other biologics in IBD, will identify the differential weighting of clinical and biochemical variables to produce biologic specific predictive models that can be readily integrated into a single predictive platform. In the current proposal the candidate therefore aims to derive novel predictive models for ustekinumab and TNF-antagonists in CD and UC, and further enhance the integration and optimization of these biologic specific models through machine learning. This personalization of biologic therapy through comprehensive predictive models which can be readily integrated into a single easy-to-use population level platform and updated routinely with emerging data (i.e. biomarkers and/or additional clinical trials) would bridge gaps in precision medicine for IBD.
Effective start/end date9/1/224/30/23


  • American Gastroenterological Association Institute (2019 Research Scholar Award 7/12/2022)


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.