Accumulating data supports the concept that IBD has a preclinical period that may be amenable to early diagnosis and intervention. Understanding the molecular underpinnings of this preclinical period in disease pathogenesis offers new opportunities in risk prediction and prevention. We have recently reported the first data from a large study using longitudinal pre-diagnostic samples from the US Department of Defense Serum repository, the PREDICTS study, where we assessed the predictive value of a panel of antimicrobial markers and a proteomic panel (>1100 proteins). Additional –omics data including glycan profiles, metabolome and exposome are currently generated on these same samples that span from up to 5 years before diagnosis to first clinical symptoms and confirmed diagnosis. We will integrate these massive multi-omics datasets to develop machine-learning (ML) based models predictive of future progression of CD and UC years before diagnosis, and to discover driver biomarkers and biological processes responsible for disease initiation. This will be achieved by constructing time-varying disease-specific networks capturing the associations across all biomarkers measured at different years before diagnosis available through the PREDICTS initiative. We will make all results publicly available via a user-friendly web-platform to facilitate the easy query, exploration and utilization of these data.