Krebstein is a project that uses generative artificial intelligence to create high-quality synthetic data for cancer research. By generating realistic but artificial patient data, the project helps researchers study cancer progression, identify rare patient groups and develop new insights, while respecting ethical and privacy constraints.

Problem

Cancer research depends on access to large and detailed clinical datasets. However, collecting and sharing this data is often limited by ethical, legal and privacy restrictions. As a result, researchers face data scarcity, which slows progress in studying disease evolution over time,  especially for rare cancers and uncommon patient subgroups.

Solution

Krebstein addresses this challenge by using generative AI models to create synthetic cancer data that preserves the statistical and biological patterns found in real datasets, without exposing sensitive patient information. The system is trained on existing clinical and molecular data, such as gene expression profiles, and can generate new synthetic samples or realistic data trajectories that simulate how cancer evolves over time. These synthetic datasets can then be safely used for advanced analyses, including subgroup identification, cancer stage prediction and biomarker discovery, supporting research that would otherwise be limited by data availability.

Success Story

• Industry: Health
• Results: The use of synthetic data enabled the identification of a previously unknown patient subgroup in medulloblastoma, a childhood brain tumor, and the discovery of gene biomarkers linked to kidney cancer stage progression.
• Impact: These results reduced the misclassification of uncertain cases in medulloblastoma and improved early-stage detection of kidney cancer, supporting more accurate diagnosis and treatment planning.