AI in Life Sciences​

Life sciences investigate living systems across multiple scales, from molecules and cells to organisms, populations and ecosystems. The SCIANCE review covers seven life science research domains represented by European research infrastructures: biobanking and biomolecular resources (BBMRI-ERIC), life science data (ELIXIR), marine biological research (EMBRC-ERIC), biological and biomedical imaging (Euro-BioImaging ERIC), early-stage drug discovery (EU-OPENSCREEN), integrated structural biology (Instruct-ERIC), and microbiology and microbial resource science (MIRRI-ERIC).

AI approaches and maturity

• Operational: Biological image analysis, bioinformatics, protein structure prediction, parts of drug discovery
• Emerging and Frontier: Multimodal data integration, predictive modelling of complex biological systems, autonomous experimentation, generative AI and AI-supported hypothesis generation

Biological & biomedical challenges

• Bridging spatial and temporal scales from molecular and cellular observations to tissue, organism and ecosystem dynamics in native physiological contexts
• Distinguishing true biological variation from technical noise
  Overcoming taxonomic, geographical and demographic sampling gaps
• Transition from descriptive to mechanistic, predictive biological models
• Identifying therapeutic compounds through target identification and early toxicology prediction

Methodological & analytical challenges

• Integration of multimodal and heterogeneous data types with different formats, scales, semantics and quality
• Interoperability and standardisation across data types, platforms and institutions
• Reconstruction, compression and representation of high-dimensional, time-resolved datasets, while preserving biological information.
• Resolving data fragmentation, incomplete metadata and variable curation quality to improve reproducibility and reuse
• Autonomous, adaptive data acquisition and experimental workflows

AI-use challenges in life sciences

• Model opacity, trustworthiness, reporting standards, reproducibility, generalisability, robustness and synthetic and AI-generated data risks
• AI-readiness of data, sustainable computational resources and federated, privacy-preserving infrastructure
• Training-data and model-output bias, ethical, legal and societal frameworks, public trust in AI-enabled research
• AI skills gaps, workforce training, retaining AI expertise in Europe and long-term AI-tool maintenance and lifecycle management

• Integration of multimodal data including imaging, omics and clinical information
• Federated and privacy-preserving infrastructure for AI-supported health research
• Biological, clinical foundation models and digital twins
• Autonomous, adaptive experimentation and laboratories
• Synthetic data generation and use for privacy-preserving AI training at scale
• AI-ready biological data ecosystems
• Trustworthy and reproducible AI services embedded in research infrastructures

Scientific progress in life sciences relies on large, diverse datasets from advanced technologies. Research is growing more data-driven, interdisciplinary, and collaborative, requiring integration across experimental methods, biological scales, and disciplines. AI is increasingly transforming several domains, but future progress will depend on interoperable and well-annotated data, sustainable research infrastructures, computational resources, harmonised standards, responsible governance, and the capacity to combine experimental, computational and analytical expertise.