The Sample Is a Sentence

by Marina VasquezHead of Semantic Systems, HMND

The pharma system that matters most right now is the one that can agree on what a sample, assay, study, or material actually is, because every handoff still breaks when the names drift and the meanings split. The newest pressure shows up in the standards work around the IDMP ontology and in fresh manufacturing ontology releases, both of which point straight at the same hard truth: interoperability only works when identity and meaning are treated as design choices, not cleanup tasks.

When three systems name one thing three ways

A lab information system, a clinical platform, and a manufacturing historian can all hold a record that looks familiar and still mean different things by it. One system thinks the entity is a specimen, another calls it a sample aliquot, and a third stores it as a material lot, and then everyone wonders why the audit trail feels haunted. That is not a formatting issue. That is meaning splintering in plain sight. The data moved, but the decision did not.

The recent IDMP ontology work is useful because it makes the medicinal product model explicit as an ontology, with precise definitions and relationships meant to support semantic interoperability across jurisdictions and systems. That matters well beyond regulatory filing. Once the ontology sits over the raw identifiers, the same term can finally mean the same thing in a submission system, a research catalog, and the downstream quality record.

The boundary that keeps biting

The real integration boundary is not the warehouse. It is the seam where source system identifiers meet governed identities. That seam runs through master data, terminology services, and the object model inside the lab and manufacturing stack, where a sample created in ELN or LIMS has to survive translation into MES, CTMS, or a regulatory package without losing its identity. If the handoff layer has no governed ontology and no identity resolution rule set, the downstream system becomes a polite liar.

Pistoia’s recent assay repository work makes the point from the other side of the lab bench. It frames in vitro pharmacology assays as structured records with controlled vocabularies and standardized fields so they can be searched, shared, and reused without turning every team into its own private language school. That is the kind of semantic layer that changes what people can decide. A scientist can compare assays without arguing about labels first, and a quality reviewer can trace the logic without reconstructing the vocabulary from email.

Governance is the actual infrastructure

Teams keep hoping a dashboard will solve what governance was meant to hold. It will not. Ontology, identity resolution, and stewardship rules decide whether a record is just stored or actually trusted. The distinction is brutal and useful. A governed semantic layer turns a pile of cross system facts into a decision surface that can survive a transfer from discovery to development to manufacturing, and the decision surface is where pharma either keeps its meaning intact or loses it in translation.

That is also why the recent manufacturing ontology work matters. NIIMBL and OAGi released biopharmaceutical manufacturing ontologies that cover process parameters, equipment, quality attributes, recipes, steps, and materials, explicitly for interoperability, analytics, submissions, and digital transformation across the manufacturing value chain. The old habit was to treat those as separate software problems. They are one semantic problem wearing different badges.

A team in a lab can get by for a while with local naming conventions. Then a study spans sites, the assay moves, the material gets repackaged, and someone has to decide whether the thing in front of them is the same entity or merely a cousin with better paperwork. Meaning has to be governed there, because nobody gets bonus points for discovering ambiguity at release time.

The best recent work is pointing in the right direction. ISO and Pistoia are building shared, machine readable representations that keep definitions, relationships, and governance close to the data itself. That is the intersection of pharma and IT that deserves attention now. Not another dashboard. A semantic layer that can carry identity across the seams where clinical, research, and manufacturing systems keep dropping it.

If this kind of handoff problem is living on your desk, hello@example.com is where the conversation starts.