Overview The analysis aimed to predict and rank individuals across various immunological and molecular outcomes post-vaccination using integrated datasets from 2020 to 2023. Robust preprocessing, feature alignment, and H2O AutoML were key to model optimization across all challenges.
Data Sources: Combined immunological (TCP, TCA) and molecular (gene expression) predictors. Modeling Approaches: H2O AutoML used in all tasks for automated model optimization; log transformations applied to stabilize outcomes. Strengths: Comprehensive feature integration, robust preprocessing, and strong validation metrics. Limitations: No clinical metadata used; constrained AutoML runtime.
Challenge 1.1: IgG-PT Levels (Day 14); Predictors: Baseline (Day 0) IgG levels.
Challenge 1.2: IgG-PT Fold Change (Day 0 to Day 14); Predictors: Baseline IgG levels from 2020–2022 datasets.
Challenge 2.1: Monocyte Percentage (Day 1); Predictors: Baseline PBMC cell frequencies.
Challenge 2.2: Monocyte Fold Change (Day 0 to Day 1); Predictors: Baseline PBMC cell frequencies.
Challenge 3.1: CCL3 Expression (Day 3); Predictors: Baseline gene expression of the KAT → STAT → CCL3 pathway.
Challenge 3.2: CCL3 Fold Change (Day 0 to Day 3); Predictors: Baseline gene expression of the KAT → STAT → CCL3 pathway.
Challenge 4.1: IFN-γ/IL-5 Polarization Ratio (Day 30); Predictors: TCP/TCA analytes and gene expression of Th1/Th2-related genes.