Trends from the Trenches: Why Semantics Matter in Life Sciences

By Bio-IT World News Staff 

August 5, 2026 | The AI boom has pushed life sciences data strategy into an uncomfortable spotlight: the answers look polished, but hallucinations, weak reproducibility, and fuzzy audit trails make them risky in regulated work. That pressure is why knowledge graphs, ontologies, and FAIR data principles are showing up in more conversations about AI readiness. In the latest episode of Trends from the Trenches, Tom Plasterer, CEO and co-founder, and Eric Little, chief data officer, of Knowledge3 break down what it takes to turn semantics into something practical. 

When a team can trace a claim back to its source, preserve provenance, and define meaning with shared semantics, it becomes possible to scale AI beyond demos and into real scientific decision making. In pharma and biotech, that matters because “facts” are often provisional scientific truths, and the cost of getting it wrong can be measured in years, budgets, and patient impact. 

A practical response is to treat semantics like a product, not a one-off model. The Knowledge3 perspective centers on closing the gap between scientific intent and usable systems by making knowledge “plug and play” through modular components and strong operational discipline. Problem decomposition breaks big business questions into smaller parts that can be delivered, repeated, and improved, instead of building a fragile, monolithic enterprise ontology.  

This is where semantic ops, or “semops,” comes in. Applying DevOps style thinking to metadata, models, and knowledge graph delivery so teams can version, test, and operationalize semantics with the same seriousness they apply to software engineering. “Context graphs” are not a new invention, though there is a renewed focus on what graphs were always meant to do: represent relationships with nuance. Context can include time, conditions, probability, and perspective, all of which are essential in biomedicine.  

The scalable way to capture that nuance is layered ontology engineering. Start with data source faithful models that mirror real tables, documents, and identifiers for one-to-one traceability, then lift into domain and subdomain ontologies (e.g. patients, diseases, trials, sites, and biomarkers) that add meaning. Finally, recompose those layers so queries can route through only what matters, supporting multiple viewpoints, such as a chemist’s receptor binding focus versus a biologist’s pathway and system effects view. The payoff is a deterministic backbone that complements — rather than competes with — AI.  

To learn more about the roles of knowledge graphs, LLMs, and cross functional friction, listen to the Trends from the Trenches podcast.