Toolchain fragmentation: why imaging, culture, and data systems need to talk to each other
Reading time: 7 minutes
A typical stem cell workflow involves imaging, cell culture, analytical measurement, data recording, environmental monitoring, and process control. In most laboratories and many manufacturing facilities, these functions are served by instruments from different manufacturers, running different software, exporting data in incompatible formats. The microscope cannot speak to the bioreactor. The bioreactor cannot speak to the quality control system. The quality control system cannot speak to the batch record.
This fragmentation is not a minor operational inconvenience. It blocks automation, prevents real-time process control, introduces transcription errors when data is moved manually between systems, and creates barriers to the regulatory documentation that clinical manufacturing requires. This article, part of the Pillar 2 series on why ancillary technologies fail, examines why integration matters and what it would take to get there.
How workflows are actually structured
The idealised picture of a stem cell manufacturing workflow is linear: thaw cells from the bank, expand them, differentiate them, characterise the product, release it. In reality, the process is iterative and branched. Cells are monitored at multiple stages. Decisions about feeding schedules, passage timing, and media changes are often made by operators based on visual assessment. Data is recorded in spreadsheets, lab notebooks, or instrument-specific databases that do not interoperate.
In a research setting, this patchwork approach is tolerable. The operator holds the workflow together through tacit knowledge and manual coordination. In a manufacturing setting, it is a liability. The GMP article in Pillar 1 described the documentation requirements for clinical-grade cell products: every step recorded, every deviation investigated, every input traceable. Meeting these requirements with disconnected instruments and manual data transfer is expensive, error-prone, and difficult to scale.
The reproducibility article in this series described how operator-dependent variation blocks commercialisation. Toolchain fragmentation is one of the structural reasons that variation persists: when the workflow depends on human judgement to bridge gaps between instruments, the consistency of the process depends on the consistency of the human.
What fragmentation looks like in practice
Several specific disconnects recur across the field.
Imaging and culture systems. Phase contrast and fluorescence microscopy are used routinely to assess cell morphology, confluency, and differentiation status. In most setups, the microscope is a standalone instrument. Images are captured, stored on a local drive, and assessed visually by the operator. The culture system, whether flasks, plates, or bioreactors, has no access to the imaging data. Decisions about when to passage, when to feed, and when to harvest are therefore reactive, based on periodic snapshots rather than continuous data. Inline imaging systems that sit within culture platforms are emerging but are not yet standard, and their output formats are not standardised across manufacturers.
Process sensors and batch records. Bioreactors used for stem cell expansion, discussed in detail in the scale-up article, generate continuous streams of data: temperature, pH, dissolved oxygen, agitation rate. This data is typically stored in the bioreactor's own software environment. Batch records for GMP manufacturing, meanwhile, are often maintained in a separate electronic or paper-based system. Linking sensor data to the batch record in real time requires integration that most facilities build manually, if they build it at all.
Analytical instruments and process control. Quality control assays, from flow cytometry to PCR to metabolite analysis, generate results in instrument-specific formats. These results inform decisions about whether a batch proceeds to the next stage or is rejected. In a well-integrated system, an out-of-specification result would trigger an automatic hold on downstream processing. In practice, results are exported, reviewed by a scientist, and the decision communicated separately to the production team. The lag between measurement and response introduces risk, particularly in continuous processes where conditions can drift during the review period.
Data formats and ontologies. There is no universal standard for describing stem cell manufacturing data. Cell culture parameters, characterisation results, and process metadata are recorded using different vocabularies across instruments, laboratories, and organisations. This makes it difficult to aggregate data across batches, across sites, or across studies, limiting the ability to apply statistical process control or machine learning to improve manufacturing consistency.
Why it matters for TechBio companies
For the company building ancillary technologies, toolchain fragmentation creates two problems.
The first is adoption friction. Your product does not enter a greenfield environment. It enters an existing workflow with established instruments, established data systems, and established operator habits. If your product requires the user to learn a new software interface, export data in a format the rest of their workflow cannot consume, or manually transfer information between your system and theirs, you have created work for the customer. In a busy laboratory or manufacturing facility, that friction is often sufficient to prevent adoption.
The second is validation scope. If your product claims to improve some aspect of the workflow, the evidence for that improvement depends on data from the rest of the workflow. A monitoring system that detects early differentiation drift is only useful if the detection triggers a response. That response requires integration with the culture system, the batch record, and the quality decision framework. If these connections do not exist, the monitoring system is an interesting instrument but not a workflow solution.
I believe integration failure as a systemic reason ancillary technologies for stem cell science struggle commercially: tools built in isolation, imaging systems that do not connect to culture platforms, data pipelines that are incompatible, manual workflows that dominate. The result is high operational complexity and poor adoption outside expert laboratories.
What interoperability would require
Genuine interoperability in stem cell manufacturing would involve several components that do not yet exist in standardised form.
Data exchange standards. The field needs agreed formats for describing cell culture parameters, characterisation results, and process metadata. Efforts toward this exist in adjacent fields, including the Allotrope Data Format for analytical chemistry and the FAIR data principles for research data management, but no stem cell-specific data standard has been widely adopted.
Application programming interfaces (APIs). Instruments need to expose their data and accept instructions through open, documented interfaces. Proprietary, closed systems lock users into single-vendor ecosystems and prevent the kind of mix-and-match integration that complex workflows require.
Middleware and workflow orchestration. Software layers that sit between instruments and coordinate the flow of data, decisions, and actions across the workflow are essential for manufacturing automation. These exist in other manufacturing industries but are not yet standard for cell therapy production.
Common ontologies. A shared vocabulary for describing cell types, culture conditions, assay results, and process deviations would enable data aggregation and cross-facility comparison. This is not a purely technical challenge; it requires community agreement on terminology and its use.
What TechBio founders can do now
Full interoperability is a long-term goal. In the interim, several practical steps can reduce adoption friction and improve your product's integration with existing workflows.
Design your data outputs to be machine-readable from the start. Comma-separated values and JSON are more useful than proprietary binaries or PDF reports. Expose your product's data through an API, even a simple one. Document your data schema and make it available to potential integration partners. Support common laboratory information management system (LIMS) integrations where they exist.
Think about where your product sits in the workflow and what information it needs from upstream and downstream instruments. Design for that information exchange, not just for standalone operation. The products that succeed commercially in this space will be those that reduce the total operational burden of the workflow, not those that add another island of functionality.
About StemCells.Help
StemCells.Help is an advisory consultancy that aids innovation and real-world impact of life science applications built on developmental and stem cell biology. Founded by Dr Paul De Sousa, it draws on over four decades of experience spanning early embryo development, animal cloning, pluripotent stem cell manufacturing, and technology commercialisation. If you build tools for these domains or work in an emerging application where the biology is the enabling technology, StemCells.Help can provide experienced scientific counsel to ground your decisions. To discuss your needs, talk to Paul.
ORCID: 0000-0003-0745-2504
Web: stemcells.help
