The Modeling Core of SCRIPT² builds on the successes of its predecessor by applying advanced machine learning and systems biology approaches to clinical and multi-omics data to model pneumonia pathogenesis and identify actionable biomarkers and therapeutic targets. A key achievement of the initial SCRIPT cycle was developing a detailed model of severe SARS-CoV-2 pneumonia, published in Nature, which highlighted unique host response mechanisms and predicted the efficacy of the CRAC channel inhibitor Auxora in mitigating prolonged critical illness.
SCRIPT² expands on this foundation, leveraging time-discretized ICU data and biological samples (bronchoalveolar lavage, nasal epithelium, blood) to create a novel latent space model of clinical states and disease trajectories. This approach integrates -omics results and clinical interventions to uncover biomarkers that predict favorable and unfavorable transitions in severe pneumonia.
The Modeling Core’s aims are:
- Latent Space Modeling: Develop an interpretable model of clinical states and transitions in severe pneumonia.
- Biomarker Identification: Identify cellular, molecular, and clinical predictors of disease trajectories.
- Model Generalization: Validate the models using external datasets.
Overall Modeling Core Schematic