A platform predicted a new use for an existing compound. That compound is now in Phase 2 clinical trials

Target-based drug discovery is still largely manual work. Developing a new drug from scratch typically takes 12 to 15 years and costs billions. Yet many approved compounds have therapeutic potential beyond their original indication that remains undiscovered simply because the analytical tools to identify it do not exist at scale. Drug repositioning - finding new uses for approved compounds - compresses the timeline to 4 to 6 years, but only if the repositioning decision can be made computationally rather than manually.

Built as part of the InnoHealth DataLake initiative with the University of Pécs, Repo-X translates the abstract language of biology into drug-disease relationships using a large-parameter AI model (50M+ parameters) trained on a multi-aspect cell model integrating 30+ big data sources and 2.25 million biomedical associations. A graph database (Neo4j) layer with word2vec embeddings and t-SNE projection creates a 2D discovery map for visual exploration. Machine learning validates predictions against known drug-disease pairs. The platform is accessible to technical users via Jupyter Hub, Neo4j Browser, and KNIME Server. After four years of R&D, Repo-X has reached TRL-6 and is being prepared for spin-off as a separate business.

Repo-X's prostate cancer drug candidate is in Phase 2 of clinical testing. The platform has moved from research to clinical reality - the most meaningful validation a drug repositioning system can achieve. Beyond prostate cancer, the work extends to systematic drug-disease prioritisation across multiple indications, opening new possibilities for pharmaceutical and medical research teams.