From FAIR Science to FAIR Manufacturing

How keeping meaning and detail attached to pharmaceutical data supports trusted reuse across R&D, technology transfer and contract development and manufacturing organizations (CDMO) operations

Somewhere in the future, a patient will receive a therapy that began years earlier with a scientific question. It By the time it gets to the patient, it has traveled through research, development and technology transfer over to manufacturing, crossing departments, organizations, sites and systems throughout.

The product moves. However, does its meaning move with it?

Pharmaceutical organizations transfer data throughout this journey. The harder challenge is keeping the knowledge and detail needed to understand it attached along the way.

Data can arrive technically complete, yet still lack what a person, system or AI application needs to interpret it. A value alone does not explain what it represents, which definition applies or under which conditions it is valid.

When Scientific Success Creates a Knowledge Problem

Within the original research team, context is often still alive. Scientists know what succeeded, what failed and why measurement matters. But that understanding lives across records, models, methods and experience.

Then something good happens: the research becomes a potential therapy. Teams across development, manufacturing and contract development and manufacturing organizations (CDMOs) must now understand knowledge created elsewhere. Success expands the audience for the evidence and creates a knowledge problem.

Files and records may move easily, but the next team still needs enough meaning and detail to interpret the evidence in its own context.

Why the Details Matter

Small details can completely change meaning. “Temperature” may be a set point or an observed value, apply to a particular process phase, use a specific unit or be governed by a limit valid only for one product and approved process version. In a bioprocess, for example, a limit applicable during growth may not apply during harvest preparation.

Separated from those details, data may remain technically correct while becoming prone to misinterpretation. As a result, teams must reconstruct context before comparing results, investigating deviations or reusing evidence. And it’s this same ambiguity that carries into analytics and AI.

Every view does not need every detail. However, relevant details do need to remain available when understanding depends on them. Connecting systems does not automatically create understanding. A common structure can improve exchange while hiding distinctions that made each source precise. Once flattened, those distinctions may be difficult to reconstruct.

Harmonize What Is Shared, Preserve What Is Specific

Semantic harmonization can preserve source detail, add semantic context and relate different worlds without flattening them. Some concepts are shared, such as batches, materials, equipment and quality events; others remain sponsor-, product-, process- or site-specific.

The goal is to harmonize what is shared, preserve what is specific and solidify what makes the relationships explicit. Teams can then see where definitions align and where differences still matter.

From Sponsor Intent to Manufacturing Evidence

A sponsor transfers more than a manufacturing recipe—it transfers product knowledge: which parameters are critical, why limits exist, how quality should be assessed and how results should be interpreted.

The CDMO executes the process through its own facilities, procedures, systems and terminology. The batch is produced in the CDMO’s operational context, while its evidence must remain understandable in the sponsor’s scientific and quality context.

Meaning therefore needs to stay attached through technology transfer, site execution, batch evidence and reporting. Vocabulary, units, relationships, rules, lineage, access controls and version history help teams understand which conditions governed the work and how its results should be interpreted.

Yield Loss Is a Symptom, Investigation Depends on the Relationships

Imagine a hypothetical batch whose yield is lower than expected. That is a signal, not an explanation. Investigators first need to identify the expectation applicable to the sponsor, product and process version.

They may then examine the material lot, equipment state, phase-specific process parameters, approved control strategy, deviations, site adaptations and applicable SOP. Each fact contributes to the investigation; interpretation depends on how those facts connect.

A semantic model makes the evidence chain explicit: what was used, what executed the process, which definition governed the assessment and in whose context the evidence applies. It supports investigation, but does not establish causality. Engineers and quality specialists still evaluate competing explanations and evidence quality.

Could AI Connect the Dots Instead?

Partially. Retrieval-augmented generation (RAG) can retrieve relevant passages using keyword, vector or hybrid search. But a manufacturing question may require evidence distributed across process definitions, equipment records, deviations and sponsor-specific specifications.

Similarity alone may not establish which definition or process version applies. A trustworthy response also needs the organization’s controlled knowledge about how terms and evidence relate.

Semantic RAG adds that organizational context through knowledge graphs, controlled terminology and semantic metadata. In Progress’ approach, these work alongside document and vector retrieval to help find evidence and the relationships needed to interpret it.

For a yield investigation, that could mean retrieving the batch result together with the applicable process version, phase-specific limits and associated deviation records. The relevant context then follows the evidence into the answer.

AI answers still require validation. Explicit relationships and provenance provide a basis for checking where evidence came from, which definition applied and why it was relevant. Specialists remain responsible for evaluating the answer and its supporting evidence.

Learn more about Progress’ semantic RAG approach

What FAIR Manufacturing Means

The FAIR principles address the findability, accessibility, interoperability and reusability of digital resources. Here, FAIR manufacturing means applying those principles to evidence as it moves from research into manufacturing.

The following is a practical manufacturing interpretation of the principles:

Findable: Discovers relevant evidence through searchable metadata and persistent identifiers, connecting the language of the person asking to the organization’s terminology

Accessible: Makes evidence retrievable through defined access methods, with authentication and authorization where needed

Interoperable: Connects different terminologies and source representations through formal knowledge models, shared vocabularies and explicit references

Reusable: Preserves relevant context, provenance, versions, usage rights and domain standards so evidence can be interpreted beyond the original transaction

These principles do not require every detail to be visible to every user; they support retaining the information needed for reuse, even when the current view or transaction does not need it.

A result can then remain connected to the definitions, conditions and history that give it meaning, even after it leaves the team that produced it.

Read the FAIR Guiding Principles

How Progress Data Platform Supports the Approach

Progress Data Platform brings together technologies for connecting enterprise information, applying semantic context and supporting AI and analytics workflows. Progress MarkLogic software provides a multimodel data foundation for complex information and relationships. The Progress Semaphore platform supports knowledge modeling and semantic enrichment, helping organizations define concepts, relate terminology and add meaningful metadata to content.

Together, these capabilities can help connect process knowledge to evidence from selected sources. Domain expertise, integration and ongoing governance remain essential to defining the models, maintaining relationships and validating how evidence is used.

The Product Moves, FAIR Helps Its Meaning Move with It

Somewhere in the future, a patient receives the therapy that began as a scientific idea. By then, it has traveled through research, development, technology transfer, manufacturing, quality review and regulatory scrutiny. Its evidence may also support analytics and AI.

Confidence in that evidence depends on whether meaning was retained, important details remained available and results can still be understood, traced and reused.

Across customers, sites and systems, a CDMO also builds confidence in the evidence behind each product. FAIR manufacturing helps preserve that trust in data by keeping meaning and relevant detail attached throughout the journey.

Context. Control. Confidence.

For life sciences organizations working to preserve context, lineage and semantic meaning across R&D, technology transfer and manufacturing evidence, Progress Data Platform helps connect complex data, documents and domain knowledge so information remains findable, reusable and trustworthy across teams, systems and partners.

If your organization is exploring how to make life sciences data more reliable for analytics, AI and evidence reuse, a personalized walkthrough can show how Progress Data Platform helps connect the knowledge that gives data its meaning.