DecodePCS 2025 – Datathon zu KI, Omics & Post-COVID: IDG RLP

Vom 04.–06. Dezember 2025 richtet das Institut für digitale Gesundheitsdaten RLP gemeinsam mit dem EPIC-AI-Konsortium der Universitätsmedizin Mainz den Datathon DecodePCS 2025 im Gutenberg Digital Hub in Mainz aus. Ziel ist es, interdisziplinäre Teams aus Forschung, Medizin und Data Science zusammenzubringen, um mit Künstlicher Intelligenz, Multi-Omics-Daten und klinischen Analysen innovative Lösungen für das Post-COVID-Syndrom (PCS) zu entwickeln. Die Teilnahme ist kostenlos, die Anmeldung ist bis zum 02. Dezember 2025 möglich.

Hier anmelden

Topic 1: AI-supported subtyping of the post-COVID syndrome (PCS) using multimodal data

Task: Development and application of AI-supported methods for identifying clinically relevant subtypes of post-COVID syndrome based on multimodal data. The aim is to identify PCS-specific patterns while considering confounding factors such as age and gender, and to characterize the subgroups with regard to symptoms. The cluster effectiveness of the multimodal dataset will be quantitatively evaluated in comparison to single-modality approaches using established performance metrics.

Topic 2: Integration of multimodal data and network analysis for characterizing PCS

Task: Integrating clinical and proteomic data to achieve a comprehensive understanding of PCS. Using multimodal data fusion and network-based analyses, the aim is to jointly explore symptom profiles and molecular signatures to identify distinct patient subgroups and their underlying biological pathways. Interactive network visualizations will be developed to link clinical features with protein-level changes and to support interpretable cluster analysis. Special attention will be given to identifying “outlier” patients—those with severe symptoms despite normal molecular findings, or vice versa—to uncover atypical biological mechanisms and refine PCS subtype characterization.

Topic 3: Using group-based proteomics information to research PCS

Task: To utilize proteomics-based groupings based on existing biological knowledge for improved interpretation of PCS data. Individual protein measurements can be aggregated into groups (e.g., functional protein sets), and suitable statistical models should be applied to investigate associations with clinical PCS manifestations. Specifically, the analysis should aim to identify what relationships exist between different symptoms (e.g., fatigue, memory problems, anxiety, depression) and functional protein groups such as those associated with inflammation, microcirculation, or neurotransmitters? Furthermore, the predictive power of such group-based omics signatures for classifying clinical phenotypes should be evaluated.