Generative AI is moving rapidly into biomedical research. Its potential goes far beyond generating text or analysing individual datasets. New models can bring together complex, multimodal health information, identify connections across different levels of biology and support researchers in exploring complex questions across large and diverse bodies of biomedical information.
For health research, this could mark an important shift. Medical knowledge is distributed across clinical records, imaging, laboratory results, genomic information and scientific literature. Generative AI offers new ways to connect these data and turn them into models that help researchers understand disease mechanisms, formulate hypotheses and ultimately develop more personalised approaches to prevention, diagnosis and treatment. However, particularly in biomedical research and medical decision-making, it is crucial that the outputs of AI models are robust, explainable and scientifically grounded. Data quality, transparency, validation, bias and regulatory requirements therefore become as important as the capabilities of the models themselves.
Two Projects, Two Perspectives on Trustworthy Generative AI
Two new projects in our Health Innovation Ecosystem illustrate how this next generation of Generative AI could take shape.
AIRIS is developing a trustworthy Generative AI research collaborator designed to help scientists investigate the complex biological processes that drive disease and shape individual patient outcomes. By combining diverse biomedical data with existing knowledge and scientific evidence, AIRIS supports researchers throughout the research process, from exploring evidence and generating and testing hypotheses to identifying potential opportunities for intervention and designing rigorous studies.
At the heart of AIRIS is a mechanism-informed approach that goes beyond statistical patterns: the platform connects molecular and cellular processes with patient-level outcomes to build dynamic, causal models of disease. Ultimately, this could help researchers understand why patients with the same diagnosis experience different disease trajectories and pave the way for more personalised approaches to medicine.
TWIN-X takes Generative AI towards patient-centred digital twins. The project integrates diverse medical information into virtual representations of individual patients, focusing on cardiovascular diseases and cancer. Its explainable and trustworthy AI model is designed to help clinicians make more informed treatment decisions.
What sets the TWIN-X model apart is its structure-first approach: rather than merging all available data into a single mathematical representation, it organises diverse data sources hierarchically into semantic units that reflect the clinical context. These units are structured around organ systems, diseases, and treatment options, offering a clinically meaningful representation of complex patient data.
Together, the projects demonstrate the breadth of emerging Generative AI applications in health, from supporting scientific discovery and understanding complex disease mechanisms to modelling individual patients and informing clinical decisions.
Building the Ecosystem Around the Technology
The next step for Generative AI in health is not simply to build more powerful models. It is to create the conditions under which they can be trusted, validated and translated into research and healthcare practice.
This requires access to high-quality and interoperable data, appropriate digital infrastructures and computing capacity, but also close collaboration between AI researchers, biomedical scientists, clinicians, technology developers, regulators and patients. Ethical and regulatory considerations need to be integrated from the outset, while validation must demonstrate that innovative models are not only technically capable but scientifically and clinically meaningful. This is where collaborative innovation ecosystems become particularly important. By bringing together expertise across disciplines, sectors and countries, projects such as AIRIS and TWIN-X can address technological development alongside the scientific, ethical, regulatory and translational questions that determine whether Generative AI ultimately creates value for research, healthcare professionals and patients.
For the EURICE Group, the two projects add a new dimension to our work at the intersection of data-driven health innovation and personalised medicine. They reflect a broader ambition across our portfolio: to help transform increasingly complex health data into trustworthy knowledge, actionable insights and, ultimately, better healthcare.