What an ELN connects to is the real buying test

Six months after an upstream cell culture run, a process development scientist needs to know why a purification step underperformed. The answer exists. It's just split between an ELN entry, a LIMS record and whatever data someone exported by hand at the downstream handoff, in systems that were never built to talk to each other.

That scene opens Drug Discovery News' October 30 report on Genedata's new notebook, and it captures where the ELN market now stands. Most labs already have one. According to the Pistoia Alliance, 81% of life sciences organizations had adopted an ELN, compared with 66% the year before. Those figures come from its 2025 Lab of the Future survey, which was released in April 2026. Once adoption is that broad, the buying question changes. It is no longer whether a lab runs an ELN but whether that ELN connects to anything else.

The scale of the disconnection is large. Industry estimates cited by Drug Discovery News put the number of discrete data silos across biopharma above 10 million, spread across aging ELNs, standalone LIMS and mismatched instrument files. Genedata, a Danaher company that sells enterprise software for biopharma R&D, is pitching a structural answer: an ELN built natively into the Genedata Biopharma Platform and linked to the molecules, samples, process parameters and assay results the platform already manages.

Biologics programs pile up the records

Drug Discovery News argues that biologics produce more disconnected records than any other therapeutic modality. One program can run through antibody screening, cell line development, upstream bioprocessing, purification and analytical characterization, generating records across five or six workflow stages, each often handled by a different informatics tool. Organizations that built their stack piece by piece often end up with several ELNs and no shared data model, sometimes within a single site.

A November 2025 survey of 150 scientists across biopharma R&D, CROs, diagnostics and pharmaceutical manufacturing put numbers on the friction. One small warning about the numbers: the survey has its own 81%, which measures something different from the Pistoia adoption rate.

Part of the reason so few scientists can work alone, according to the survey write-up, is that finding and reusing earlier work across disconnected systems has become a technical job in its own right. For a team that measures its informatics stack by uptime or user counts, the import-export share is probably the more honest benchmark of whether the systems actually help.

Shared objects instead of API bridges

Genedata's central claim is architectural. Rather than being a bolt-on module wired in through APIs, the ELN sits inside the platform's data model. Experiment records therefore use the same native objects as assay results, sample inventories and molecule registrations. A cell culture result entered by a scientist connects directly to the matching sample lineage, the upstream process parameters and the downstream assay outputs the platform already holds. No separate export is needed.

The argument laid out in Drug Discovery News is that standalone ELNs joined by APIs or middleware need data mapping at every exchange point, which brings back the manual burden the integration was supposed to remove. In a biologics program with five or six stages, every one of those handoffs becomes a mapping point that has to be maintained.

A LIMS-versus-ELN guide from LIMSey sets out that distinction clearly, and it explains why the problem has persisted for so long. Notebook entries are meant to be free-form, since an experiment exists to discover something. Sample records are meant to be rigid, since they must stay comparable over many years. Connecting the two is not a new idea. Drug Discovery News cites a 2013 framework in the Journal of Laboratory Automation, which argued that an ELN is most valuable once it is structurally tied into LIMS, chromatography data systems and scientific data management systems.

LIMSey sums up the split in one line: an ELN answers "what did we learn?" and a LIMS answers "where is it, and did it pass?" The purification question from the opening needs both answers at once. The scientist wants to know what was learned in the culture run and also exactly which material moved downstream, which is why the join between the two record types is where the time goes.

Each side also brings its own record-keeping obligations. According to LIMSey, a LIMS logs every change along with the equipment used on a test and the calibration it was under. ELNs in regulated life-science work carry FDA 21 CFR Part 11 controls such as audit trails and electronic signatures, plus signing and witnessing that serve as dated evidence for intellectual property. A shared data model would need to keep both sets of controls intact, so that is worth testing alongside the links themselves.

There is a serious counterargument. LIMSey argues that a research group is usually better off with its own ELN than with one system bent to do two jobs, although it makes that case for manufacturer test labs and accepts the ELN-plus-LIMS model for drug discovery.

Coverage from antibody design to release

Drug Discovery News reports that the Genedata ELN covers biologics development end to end, starting at molecular design and running through analytical release. Its workflow coverage includes:

• Antibody discovery, candidate engineering, and expression development, grouped under molecular biology and screening
• Cell line development, bioprocess development, and purification, grouped under upstream and downstream processing
• Analytical development, formulation, and stability workflows, grouped under formulation and analytics
• External collaboration: laboratory workflows that span CROs, CDMOs, and internal labs
• Complex modalities: cell and gene therapies, ADCs, and bispecific antibodies are supported
• Instrument connectivity: laboratory instruments link to the platform directly

The first three groups line up with the stage sequence that makes biologics so record-heavy, from screening through cell line work, purification and characterization. That alignment is what the shared-model argument depends on. A notebook that covers only some of those stages would leave the remaining handoffs to the export routine the survey respondents complain about.

The external collaboration line deserves the most attention. How much it helps in practice will depend on how each partner's work is captured, which is worth checking in any evaluation.

Shadow AI and the thousand-dollar rerun

The 45% of surveyed scientists using personal generative AI accounts are a signal about demand. Rob Brown, global vice president and head of the Scientific Office at Sapio Sciences, told The Scientist in March that ELNs are still quite passive. A scientist often goes into the ELN to find what happened in a previous experiment, leaves to talk to colleagues about what to do next, then comes back to document the next experiment. When organizations don't provide AI tools, he said, researchers bring their own.

Brown was discussing a Sapio Sciences survey of 150 scientists across large pharma and biotech, spanning discovery, development and manufacturing, mostly in the US and Europe. Many scientists, The Scientist noted, now see ELNs as little more than digital filing cabinets. With AI, Brown said, people want something else: an active helper in their thinking, closer to a co-scientist that helps them decide what to make and test.

He was clear that it is better for the organization to supply those tools and put guardrails around the infrastructure. But scientists know what the leading edge looks like, and if an organization isn't keeping up, he said, they will find ways to do it themselves. For informatics leads, that makes the 45% figure a governance question as well as a productivity one.

Brown also said some experiments get repeated because there isn't a good way to find out they've already been done, and each unnecessary repeat costs a thousand dollars or more. A research informatics capability, he said, should tell scientists whether an experiment has been done before, or something similar, and let them simply ask, "Has anybody done this experiment before?"

The friction he describes is specific. Finding out is possible in theory, but if it means asking a data scientist to run the search or building a long, complicated query, that is a barrier. And the harder the question is to answer, he said, the more likely someone is to say, "Let me just do it again." Read next to the survey's 5% figure, that barrier is the normal case. Most scientists cannot get to their own results without specialist help, so a rerun can end up looking like the faster option.

That only works if the records underneath are connected, which is Drug Discovery News' point about AI: these tools need clean, linked data, and fragmented ELN records break that requirement at the source. Genedata's own November 3 note on the coverage makes the same pitch, describing the shared model as a unified data foundation for scientific workflows and AI-driven R&D.

Where old records fit in

For a site running several notebooks with no shared model, the scenario Drug Discovery News describes, the practical question is which historical records get brought into a unified platform and which stay where they are.

SciSure, the ELN and LIMS vendor formed when eLabNext and SciShield merged, makes the mirror-image case for start-ups. Set up an ELN and sample management before the first experiment, it advises, and agree on freezer layouts, box positions and sample naming up front to avoid a painful migration of hundreds of existing records later. Even three full-time scientists, it notes, can generate dozens of samples within months and eventually hundreds.

An established biologics site is past that window by definition. SciSure warns that adding sample management late means reconstructing missing information and renaming records while experiments keep running, which suggests the scope of migration could cost a consolidation project more than any gap in the feature list. Its point about protocols applies here as well. Founding scientists often run the same experiment in different ways, and several unconnected notebooks could hold that same divergence in their records.

Genedata's approach could matter most for teams whose programs span several biologics workflow stages, run work through CROs and CDMOs, or handle complex modalities such as ADCs or bispecific antibodies. A lab running one assay workflow at one site has fewer handoffs to protect, and a well-kept standalone notebook may cover what it needs.