BLOG
AI in life sciences may be evolving fast, but true progress requires more than powerful models. It demands data that is structured, connected, and trustworthy. During this year’s BioTechX Europe, ONTOFORCE CEO Brecht Claerhout hosted a panel with leaders from MSD, Novo Nordisk, and Roche to explore how knowledge graphs are shaping the future of AI across pharmaceutical research and development.
Speakers:
Bryn Roberts, SVP & Global Head of Data, Analytics & Research, Roche
Nicklas Walldorf Gaardsted, Director, Head of Data & Modelling, Novo Nordisk
Guglielmo Iozzia, Director – Machine Learning/AI and Applied Mathematics, MSD
“All organizations have climbed the AI ladder fast,” said Gaardsted from Novo Nordisk. “But building the right foundation is crucial for sustainable progress.”
That statement captured the essence of the discussion: the future of AI isn’t just about acceleration, it’s about maturity. In this blog, we’ll cover some of the topics and themes that the panel discussion touched upon, such as building a strong foundation for AI, human oversight, preparing for multimodal intelligence, and the role knowledge graphs play in all of this.
It bears repeating: high-quality, trustworthy data is the lifeblood of effective AI. Without it, even the most advanced algorithms produce unreliable or misleading results. Trustworthy data ensures that AI systems learn and operate from accurate, representative, and context-rich information.
Establishing the accuracy of data is still an area that pharma organizations are struggling with. MSD’s Guglielmo Iozzia described how teams spend valuable time validating whether information is correct. It’s clear that data lineage and traceability remain critical, not only to help reduce manual and time-consuming efforts to validate data but to also ensure compliance. Throughout many stages of the drug development lifecycle, the ability to audit and track every data point is non-negotiable for compliance and trust.
Reliable data, the panel agreed, is the first building block of intelligent generation and automation.
As organizations strengthen their data quality, the next step is connecting this high-quality data. There are a few ways to approach this. Roche’s Bryn Roberts highlighted how knowledge graphs can serve as the connective tissue for trustworthy AI, providing depth, context, and the ability to detect inconsistencies or hallucinations in GenAI outputs.
Roberts shared an example from Roche in which they used a knowledge graph to improve accuracy of an AI-powered application that summarized patient notes. The interplay between AI and knowledge graphs supports more reliable models, but do knowledge graphs also stand to benefit from AI? The panel members believe so. Gaardsted from Novo Nordisk added that in turn, GenAI can help curate and enrich relationships between data elements — deepening a knowledge graph and amplifying its value.
While AI-powered automation continues to advance, all panelists agreed that humans must remain firmly in the loop. Roberts stressed that every AI-generated assertion must be explainable and reviewable, transparency isn’t optional. Human oversight ensures reasoning chains are traceable and ethical considerations are upheld.
This represents a kind of collaboration between humans and AI agents: one where algorithms assist, but scientists guide. Governance, lineage, and critical review are thus not barriers to progress, they’re actually the enablers of sustainable AI.
Beyond compliance, human oversight also plays a crucial role in building trust and accountability. In a field where a single decision can impact patient outcomes or regulatory standing, scientists must understand not just what an AI concluded, but why. This requires systems that make reasoning visible and reproducible.
The panelists emphasized that achieving this level of transparency depends on solid data governance and clear provenance. When every data point is traceable to its origin and context, it becomes possible to confidently evaluate AI-driven outputs. As life sciences organizations operationalize AI, these governance frameworks will serve as the backbone of trust.
The next chapter of AI in life sciences lies in multi-modality, meaning the ability to integrate and interpret diverse data types such as omics, text, imaging, and clinical data within unified analytical frameworks. By bridging these modalities, organizations can gain a more holistic understanding of biological systems, disease mechanisms, and treatment outcomes.
However, achieving this vision requires more than simply combining datasets. Each data type comes with its own structure, standards, and context. To make them interoperable, they must be anchored in a shared semantic layer that defines how entities relate to one another across domains. This is where knowledge graphs play a central role.
Knowledge graphs provide the connective framework that allows multi-modal data to “speak the same language.” They harmonize terminologies, preserve provenance, and capture the intricate relationships that exist between data concepts. In doing so, they create a semantic backbone on which AI can reason effectively. Potentially not just correlating data points but enabling an understanding of the biological context that links them.
Both MSD and Roche emphasized that multi-modal intelligence can only emerge from strong, connected data foundations. Without shared semantics and interoperability, integration remains out of reach — and AI insights risk remaining shallow or fragmented. But with knowledge graphs underpinning these efforts, organizations can unlock truly contextualized, explainable, and high-value insights.
As the industry moves toward this new era, the combination of multi-modal data and knowledge graph-driven connectivity will define the next generation of discovery, where AI doesn’t just analyze information, but understands it.
As the panel made clear, the future of AI in life sciences will be defined not by how fast organizations adopt new technologies, but by how well they ground them in high-quality, connected, and explainable data. Knowledge graphs, governance, and multi-modal integration are not abstract ideals. Rather, they are the essential building blocks for successful, sustainable, and trusted AI innovation.
At ONTOFORCE, we see this evolution reflected in how organizations use DISQOVER to transform fragmented data into connected intelligence. Built on robust knowledge graph technology, DISQOVER enables teams to explore relationships across diverse data types through an accessible, intuitive interface. Its powerful export and integration capabilities ensure that insights don’t remain locked in the platform but can flow directly into research, analytics, and downstream workflows. In doing so, DISQOVER helps organizations climb the AI maturity ladder with confidence.
