AI/ML and Emerging Methods represent a rapidly expanding cross-field area within ELIXIR CZ, providing foundational computational capabilities, advanced analytical methods, and shared expertise that support all scientific domains of ELIXIR CZ. As life sciences increasingly rely on machine learning—from structural predictions to multi-omics integration, image analysis, natural-language processing, and drug discovery—AI readiness becomes a critical component of national and European research infrastructures.
This area connects ELIXIR CZ with the broader European AI ecosystem, including ELIXIR’s Machine Learning Focus Group, the European AI Factories, AI for Science initiatives, and national activities under CZAI (Czech Academy of AI). It leverages the GPU-enabled compute capacities of MetaCentrum, CERIT-SC, IT4Innovations, and the national AI Factory, as well as secured environments for sensitive data analytics in collaboration with GDI, FEGA, and the emerging national TRE initiatives.
The aim for 2026–2030 is to develop ELIXIR CZ into a coordinating hub for AI in Czech Life sciences, providing tools, methodological guidance, secure execution environments, curated datasets, and training that equip researchers to responsibly and effectively deploy AI methods.
Current Situation and Identified Strengths
Artificial intelligence methods are now embedded across all strategic areas of ELIXIR CZ — from structural bioinformatics and chemical biology to genomics, data stewardship, and human data analytics. Their rapid adoption has led to the emergence of an expanding national community of researchers and developers with practical expertise in deep learning, machine learning, natural-language processing, multimodal modelling, and data-driven scientific workflows. This expertise is distributed across multiple scientific domains, creating a robust foundation for building a coordinated cross-field AI programme.
We can present multiple success stories of the application of AI within ELIXIR CZ tools and services. For example, AlphaFold database uses Mol* as a visualization tool. Furthermore, ELIXIR CZ provides multiple AI-based tools in enzyme engineering, structural bioinformatics, and other fields. Finally, ELIXIR CZ is in touch with quantum computing.
ELIXIR CZ benefits from exceptionally strong technological and compute infrastructure, provided through long-standing collaboration with e-INFRA CZ partners (CESNET, CERIT-SC, IT4Innovations). Researchers have access to state-of-the-art GPU-enabled HPC nodes equipped with NVIDIA L40s, A100, and H100 accelerators — a level of hardware maturity comparable to leading European AI centres. These systems support both training and inference for modern AI models, including large transformers, graph neural networks, diffusion models, and advanced image-analysis algorithms. Importantly, the infrastructure already supports secure deployment of proprietary models, enabling controlled analysis of sensitive datasets.
Thanks to integration with MetaCentrum, users can run a range of pre-installed life-science AI services — including AlphaFold, structure-prediction workflows, and numerous ML libraries — and also perform custom development on the latest GPU cards. The national AI Factory further extends these capabilities by offering large-scale compute allocations suitable for demanding workloads, including generative models and multimodal architectures. With these resources, ELIXIR CZ is well positioned to train, optimise, and deploy advanced domain-specific AI tools.
A major strength of the ELIXIR CZ is its close interaction with data management and FAIRification efforts. AI has enormous potential to improve metadata quality, automate dataset annotation, and support semantic enrichment of diverse biological datasets. Conversely, AI methods depend critically on well-curated, interoperable, machine-actionable data. This mutual dependency creates natural synergies with the DSW/FAIRification programme, which positions ELIXIR CZ as a site capable of producing AI-ready FAIR datasets aligned with European standards.
The cross-field AI programme is also closely aligned with the Human Data activities of ELIXIR CZ. Future applications of AI in clinical genomics, imaging, and multimodal analytics will require secure execution environments—Trusted Research Environments (TREs)—capable of supporting GPU-accelerated ML workflows while protecting sensitive data. The Sensitive Cloud provided by e-INFRA CZ already offers a foundation for such environments, and collaborations with European AI Factories and GDI/FEGA initiatives will enable deployment of federated AI pipelines and privacy-preserving analytics.
Internationally, ELIXIR CZ is an active member of the ELIXIR Machine Learning Focus Group, contributing to European guidance on explainability, reproducibility, benchmarking, and responsible ML. These connections ensure that ELIXIR CZ is both aligned with and contributing to emerging continental standards for AI in life sciences.
In summary, ELIXIR CZ enters the 2026–2030 period with a strong combination of computational infrastructure, expert community, integration across scientific domains, synergy with FAIR data stewardship, and alignment with secure environments for human data. These assets provide an exceptional starting point for building a robust, European-aligned AI ecosystem within the Czech life-science community.
Challenges and new directions
AI in life sciences is evolving faster than traditional computational infrastructure and governance models can adapt. The next period demands purpose-built environments, better integration with FAIR data systems, and responsible use of increasingly powerful generative models.
- A central challenge is that AI-ready data does not exist “by default”. Many biological datasets require extensive curation, harmonisation, semantic annotation, and privacy protection before they become usable for training or inference. ELIXIR CZ must work closely with data stewards, curators, and domain experts to build high-quality, machine-actionable datasets that respect FAIR and privacy principles.
- A key challenge is the need for secure AI execution for sensitive data, including genomic, clinical, and histopathology data. As generative models and multimodal transformers become standard tools, their deployment must occur in TRE-compatible environments with auditable workflows, containerised execution, and restricted outbound communication. ELIXIR CZ will need to develop or adopt such environments in alignment with European FEGA and GDI developments, as well as national TRE efforts at 1st Faculty of Medicine.
- AI/ML tools require sustainability: continuous re-training, monitoring for model drift, reproducibility frameworks, and governance for model sharing. Without dedicated infrastructure, large-scale models—especially multimodal or LLM-based—cannot be deployed effectively in biological contexts.
- Finally, researchers increasingly expect user-friendly interfaces, standardised pipelines, and community-driven training. The challenge is to translate cutting-edge ML methods into accessible and well-supported services that can be used across diverse life-science disciplines.
This period therefore calls for ELIXIR CZ to move beyond simply providing compute resources, toward a holistic European-style AI ecosystem: secure, reproducible, FAIR, interoperable, and domain-aware.
