InnoGraph Bio: From Biomedical Questions to Verifiable Evidence

Biomedical knowledge is vast, interconnected, and difficult to navigate. InnoGraph Bio, a research demonstrator developed by ICI Innolabs, explores how artificial intelligence can make this knowledge easier to investigate—without turning the AI model itself into an unquestioned source of truth. The demonstrator combines a bounded AI agent with PrimeKG, a biomedical knowledge graph connecting diseases, genes, drugs, biological processes, phenotypes, and other medically relevant concepts. A user can pose a natural-language question, and the agent translates it into controlled graph-retrieval operations. The resulting paths are then checked independently before being presented as structured evidence. For scientists, this approach demonstrates a transparent architecture for agent-assisted knowledge-graph exploration. Retrieval tools, graph data, verification logic, and evidence artifacts remain separate components, making each step inspectable and reproducible. The system also models failure explicitly: unsupported requests may be rejected, incomplete evidence is labelled accordingly, and unsuccessful checks are not silently converted into confident answers. For a wider audience, InnoGraph Bio offers a glimpse of a more responsible form of scientific AI—one that does not merely generate plausible explanations, but shows how an answer was assembled and whether the underlying evidence passed verification. InnoGraph Bio is currently a TRL 4 research demonstrator. It is intended for experimentation and scientific exploration, not clinical diagnosis or medical decision-making. Future development will focus on broader validation, expert evaluation, stronger provenance, additional biomedical resources, and testing in operational research environments. or explore its scientific method and development pathway.

Try the InnoGraph Bio online demonstrator or explore its scientific method and development pathway.