Earlier this month I attended BioTechX Europe 2025 in Basel (6–8 October), an event that brings together thousands of professionals across pharma, biotech and technology. The conference spanned everything from precision medicine and drug discovery to quantum pharma, bioinformatics and lab automation. I’ve focused on the sessions that really resonate with our work in technology and data: AI in clinical trials, analytics platforms and LLM-driven knowledge retrieval.
This perspective is shaped by our journey at Acuity Knowledge Partners, following the acquisition of Ascent, where we’re bringing together deep domain expertise with engineering and data science capabilities. Our goal is clear: invest globally in pharma AI and data platforms and apply proven practices from other industries to solve real business problems in life sciences.
Here are some of my highlights and learnings from the event.
What shaped the conversation.
One of the standout sessions for me was Benjamin Gmeiner’s (Novartis Pharma) talk on “Digital Twins for Clinical Trials”, alongside the panel “Digital Twins in Drug Development: Accelerating Clinical Trial Timelines”, with AstraZeneca and Novartis. Both explored how simulated control groups could replace traditional placebo arms; a particularly valuable approach for rare disease studies where patient recruitment is challenging.
Key learnings:
Digital twins can significantly reduce patient burden and speed up recruitment in clinical trials.
Regulatory acceptance is growing, but transparency and auditability remain essential.
Successful adoption requires close collaboration between data scientists, clinicians, and regulatory teams.
Trust in AI and Data Platforms.
Trust and governance were front and centre in the keynote “Bridging the trust gap: making AI work for R&D” by Stuart Whayman (Elsevier), and in the panel “Designing Data Platforms to Enable AI to Revolutionize R&D” with leaders from Sanofi, EPAM, Bayer and AstraZeneca.
Key learnings:
Data quality and governance are the foundation for trustworthy AI—model performance is secondary if the data is flawed.
Cross-functional teams must define clear accountability and data lineage to ensure compliance and reproducibility.
AI should be used to stress-test designs and accelerate insight, not to replace human judgement in critical decisions.
LLMs, Knowledge Graphs and Agentic AI.
Sessions led by Alexander Jarasch (Neo4j) and Tankred Ott (Novo Nordisk) and the “Are LLMs enough?” panel (Bayer AG, SAS, MSD, Roche), demonstrated the power and limitations of LLMs in regulated environments.
Key learnings:
Combining knowledge graphs with LLMs (e.g., Graph-RAG) improves answer relevance and reduces hallucinations.
Agentic AI can automate document review, flag inconsistencies and support compliance in regulatory submissions.
LLMs should act as orchestrators of tools and workflows, not as autonomous decision-makers.
AI Use Cases in Drug Discovery: Microsoft’s Perspective.
Thomas Balkizas (Microsoft) presented a practical overview of AI use cases in drug discovery, illustrated by a comprehensive slide of “hero workloads”.
Key learnings:
AI is already delivering value in compound safety, synthesis, target discovery and lead optimisation.
AI-driven biomarker discovery and clinical/virtual trials are accelerating research and improving cohort selection.
Lab automation and generative chemistry are making R&D more efficient and data-driven.
Each use case was tied to a clear business outcome—improving decision-making, reducing cycle times and enabling more reliable, data-driven research. This reinforced the message that AI’s value is measured by its impact on business problems, not just technical performance.
Clinical Data Platform.
When it comes to data platforms, the pharma industry has explored wide and deep on AI, data integration, and governance to enable scalable, compliant, and insight-driven research.
Key learnings:
Composable architectures, as seen in Bayer’s approach, enable flexible and secure deployment of AI/ML across diverse workflows.
FAIR principles, championed by Roche and AstraZeneca, are foundational for GenAI readiness, improving data quality, traceability, and compliance for regulated environments.
Knowledge graphs and semantic layers (Graphwise, Crown Point) transform data platforms into engines for contextual analytics, while agentic AI (Elsevier) streamlines evidence synthesis and workflow orchestration.
Unified catalogues and self-service platforms (AstraZeneca, Bayer) further empower scientists, reduce manual effort, and accelerate insight generation—ultimately supporting faster, safer, and more reliable R&D.
The Expo Floor: Platform Partners and Their Messages.
These themes were brought to life on the exhibition floor. Microsoft focused on responsible AI and knowledge graph-driven retrieval, emphasising that LLMs should act as orchestrators, not decision engines. Snowflake showcased its Health Data Cloud as the backbone for secure collaboration and governance-first AI pipelines. AWS highlighted scalable infrastructure for trial simulation and analytics, with a strong push on cost optimisation and multi-modal data integration.
These aren’t just technical updates—they’re signals of where the industry is heading: data contracts, lineage and compliance-ready architectures as the foundation for AI that matters.
Our Commitment Post-acquisition.
With Ascent now part of Acuity Knowledge Partners, we’re doubling down on pharma. Our approach is simple:
Start with the business problem—shorter timelines, fewer deviations, better compliance—not with the model.
Invest in data quality and governance as the bedrock for every AI initiative.
Co-create solutions with sponsors and CROs, blending domain expertise with engineering rigour.
We’re excited by trends like digital twins, Graph-RAG and analytics engineering. We’ll keep learning from domain experts to make these ideas real and valuable.
Closing Thoughts.
BioTechX 2025 reinforced a simple truth: progress in pharma AI depends on data quality, governance and solving real business problems. If we stay curious, listen to clinicians and statisticians, and build platforms that respect regulatory constraints, we can turn these innovations into everyday practice—and move the needle where it matters most.
Xinye Li
Head of Data Science & BI
