Core Projects

INF: Data Management, Information Infrastructure, and Application

Summary

The CRC1709 aims to integrate multi-dimensional data from compatible shared disease models regarding myeloid plasticity. For efficient and sustainable data management, a central data integration hub is established for the CRC. On this hub, all data relevant for the project are collected and integrated to allow for analysis and modelling across all data sources.

State of the Art

Challenges in Biomedical Data Management: Biomedical research often struggles with integrating diverse, sensitive, and high-volume datasets across institutions.
Sensitive Patient Data: Compliance with GDPR and ethical standards is critical for handling clinical data.
FAIR Principles: Modern research demands that data is Findable, Accessible, Interoperable, and Resusable for long-term usability.
Advancements in AI: Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) enable structured information extraction from unstructured data sources.
Unexplored Potential: Current research underutilizes rich data from electronic health records (EHR) and sequential clinical analyses.
Need for Integration: Harmonizing clinical and experimental datasets is essential for advancing translational research and understanding cellular plasticity in cancer.

Preliminary Work

Clinical Data Warehouse of MED V: Centralized repository for structured data
Medical Data Integration Center (MeDIC): Mature infrastructure ready
SMART-CARE-Linked Data Repository: Harmonizing diverse datasets
Portal of Medical Data Models: FAIR-compliant portal for annotated data
Data Analysis Pipelines in Myeloid Leukemia: Genomic / clinical data tools
Standardized Data Workflows: Established SOPs ensuring interoperability
Experience with FAIR Practices: Implementation of data sharing
Multimodal Omics Integration: Expertise in diverse data types
HPC Collaboration: Leveraging high-performance computing
AI Tools for Data Extraction: Deployment of LLM-based pipelines

Specific Aims

Working in Progress

(1) Multimodal data integration with extended EHR annotation

Integrating diverse omics data (scRNA-seq, scATAC-seq, bulk RNA-seq, Ribometh-seq, proteomics, metabolomics)
Advanced tissue profiling: Spatial multiplex imaging and multi-color flow mytometry (MFC)
Incorporating biobanked patient samples, standardized PDX mouse models, and genetically engineered disease models
Leveraging high-dimensional patient data from clinical data warehouses and external sources
Establishing a unified platform for cross-project data accessbility and analysis

(2) Cloud-enabled data analysis pipelines for reproducibility and scalability

Developing containerized bioinformatics pipelines for standardized and reproducible analyses
Utilizing High-Performance Computing (HPC) for large-scale omics and imaging data
Ensuring scalability across local and cloud-based infrastructures

(3) Automatic extraction of structured information from unstructured reports

Designing and deploying LLM-based tools for extracting structured clinical and research data
Creating a local data warehouse populated with structured outputs from EHRs and free-text sources
Enabling “talk-to-your-dataset” functionalities for seamless interaction with unstructured data
Facilitating predictive analytics and research-ready data pipelines with privacy-compliant LLMs

Collaborations

The INF project will collaborate with all CRC projects. A typical example for data integration is project A06, which processes PDX and patient samples; data modalities include whole genome sequencing, RNA-sequencing (scRNA-seq CITEseq, bulk RNA-seq), proteomics, methylation arrays and imaging data with an estimated data volume of 10 GB per sample. Project B02 specified similar data integration requirements. Specific data integration needs were also specified, for example project A07 plans scTCR-seq with approximately 260 GB per sample; or project B04 will process scATAC-seq, imaging and amplicon sequencing with an overall raw data volume of 1 TB.

PRINCIPAL Investigator