MINGYUE XUE

Core Project: Global GEM Project: https://crohnsandcolitis.ca/Research/Funded-research/The-gem-projec

World’s Largest Prospective Crohn’s Disease Cohort (107 international sites in 7 Countries, $22 million CAD invested)

RWE Study Leadership:

• Directed end-to-end RWE generation for 5K+ participants: SAP design → data acquisition → HPC execution.
• Managed 80% team workflow covering EMR, multi-omics (proteomics/metabolomics), and registry data.

Advanced Analytics & Integration:

• Statistical Modeling Pipeline:
• Applied CoxPH with restricted cubic splines to characterize non-linear exposure-disease relationships (e.g., diet-CD study).
• Identified key predictors via LASSO, Elastic Net, and causal inference techniques.
• Machine Learning Pipeline:
• Developed supervised models (XGBoost, SVM, DNN) for disease prediction; optimized via grid search hyperparameter tuning.
• Achieved AUC=0.89 for preclinical CD prediction – highest reported performance.
• Multi-Omics Integration:
• Standardized variables using ICD-10 and medical ontologies; harmonized EMR, registry, and multi-omics data.
• Processed shotgun metagenomics (HUMAnN3, MaAsLin3) from raw FASTQ to taxonomic profiles.
• Translational Impact:
• Built interactive R Shiny/Power BI dashboards for cohort monitoring, reducing reporting time by 40%.
• Collaborated with medical affairs and public health agencies to translate analytics into strategic insights.
• Delivered 4 oral presentations at DDW/ECCO (top 5% selection); mentored 10+ junior researchers.

• Predictive Analytics: Developed LR model (AUC=0.85) on 50K+ EMR records, reducing HF readmissions by 12%.
• Process Innovation: Automated sepsis detection via Python/SQL, cutting analysis time 80%.

• Flagship Project: Regression-Driven Risk Stratification Initiative (10M+ Records | Provincial Policy Integration)
• Advanced Statistical Modeling:
• Led multivariable logistic regression (AUC=0.92) and CoxPH with splines to model diabetes/NAFLD risk, identifying critical thresholds (HbA1c≥6.5%, HR=4.2).
• Causal Inference & Policy Translation:
• Built propensity score matching pipelines (STROBE-compliant) to establish evidence for screening guidelines covering 3.5M high-risk residents.
• Technical Implementation:
• Developed Shiny dashboards for 3 county health commissions; optimized feature selection for skewed clinical data.