A framework designed for discovery
Overview
This project integrates machine learning, multi-omic profiling, and causal inference methods to build and validate a comprehensive predictive framework for IBD flare prediction and disease course modeling. The approach combines large-scale real-world clinical data with molecular datasets to move from prediction to mechanistic insight and therapeutic target discovery.
Experimental / Computational Methods
Machine learning model development using EMR data from over 4,000 IBD patients at Sheba Medical Center, with treatment switching as a flare proxy; integration of microbiome and metabolomics data to identify molecular markers associated with flares; and application of causal inference methods to assess the role of specific factors in flare pathogenesis.
Data Sources / Models Used
Electronic medical records from 4,000+ IBD patients at Sheba Medical Center, molecular datasets including microbiome and metabolomics profiles, and integrated EMR-omic datasets analyzed using state-of-the-art causal inference and machine learning frameworks.
Analytical / Translational Focus
Development of a validated, clinically deployable prediction and recommendation system for personalized IBD flare prevention, identification of novel metabolite-based therapeutic targets, and discovery of causal mechanisms underlying IBD disease course — with translational implementation anticipated within 3–5 years, contingent on additional funding for data generation and development.
Powering the science
Yael Haberman, MD PhD, Colton Consortium Member
Professor, Gray Faculty of Medical and Health Sciences, Sheba Medical Center, Tel Aviv University
Elhanan Borenstein, PhD, Colton Consortium Affiliate
Professor, School of Computer Science and AI, Tel Aviv University