TOOL-08: Towards frontier AI for healthcare

Call goal: build a pan-European community and a roadmap for the next generation of frontier AI models in healthcare. Proposals must identify and curate multimodal datasets (including omics), plan privacy-preserving use through the European Health Data Space (EHDS), and map how to train and evaluate such models on EU AI Factories.

What we contribute:

·       Practical expertise in curating, harmonizing and annotating genomic and multi-omics datasets for AI training.

·       Lessons from building and validating clinical AI (GAIC), feeding the roadmap's safety and efficacy review.

·       Input on combining omics with clinical data in multimodal foundation models.

·       Industry perspective on deployment, regulation (AI Act) and scalable AI infrastructure.

·       Hands-on experience with the full range of genomic data, a frontier model must cover: short and long reads, structural and non-coding variants, and native methylation.

·       Practical knowledge of the security and privacy standards (ISO/IEC 27001, ISO 27799, GDPR) needed for privacy-preserving data use through the EHDS.

·       Experience exposing genomic analysis through APIs and integrating with clinical systems, informing how frontier models could be deployed in practice.

Why it strengthens the consortium: the roadmap must rest on what it takes to build and deploy clinical AI on real omics data. Geneyx adds that practical industry view: curating and harmonizing genomic and multi-omics datasets for model training, lessons from developing and evaluating a clinical AI classifier, and first-hand experience, as an ISO 13485-certified company pursuing CE-IVDR certification, with the regulatory and deployment questions the call asks the consortium to assess. This complements academic and infrastructure partners and widens the multidisciplinary stakeholder base the call requires.