RNA Handoffs Need an Identity Layer

by Marina VasquezHead of Semantic Systems, HMND

Novartis and Abogen’s agreement puts useful pressure on an old fiction: that a promising RNA construct can cross from an originating laboratory to a global development organization through a well assembled folder. The deal gives Novartis worldwide rights to ABO2203, an mRNA encoded CD19 by CD3 T cell engager, along with options on additional programs from Abogen’s RNA platform. Abogen is due $$575 million upfront, with potential payments reaching about $$7.2 billion if development, regulatory, and commercial conditions are met.

A folder can carry files, but it cannot carry meaning.

The entity that should exist is the RNA Development Identity Authority. Most collaborations lack one. Each organization keeps its own systems, naming habits, access controls, and carefully cultivated suspicion of the other side’s spreadsheet. Then a construct leaves discovery, and everyone discovers that “the molecule” was a sequence record in one place, a formulation candidate in another, and a manufacturing reference standard somewhere else.

This is how a promising handoff becomes a decision problem.

The handoff begins before the license

At Abogen, a construct may have acquired an internal identifier during design, another during screening, and a third when a formulation was selected for animal studies. Novartis may receive a package organized around an asset number. Manufacturing may identify the same material by a batch number. The analytical group may use an assay specific identifier that points to a method version and a plate map.

All of these identifiers can be locally correct. Together, they can still fail to identify the same thing.

Semantic interoperability gives each identifier a defined place in a shared model. The construct has a stable identity linked to its sequence, chemical modifications, intended product, and design lineage. The formulation has its own identity linked to composition, concentration, container, and preparation procedure. A sample points to the material from which it came, the operation that created it, and the conditions under which it was handled.

This distinction becomes important when a result travels. A potency value without the assay version, reference material, sample identity, and calculation rule can look complete in a data transfer while remaining scientifically ambiguous. A concentration reported against a different standard can quietly turn an apparent improvement into a measurement artifact.

If two teams cannot agree what a sample is, they have a meaning problem pretending to be infrastructure.

A file transfer cannot preserve a decision

Picture a method transfer meeting at 08:30, with discovery scientists on one side and analytical development on the other. The originating team presents a potency assay that supported candidate selection. The receiving team has the protocol, raw data, and summary report. Everyone has received the files. The trouble arrives when someone asks which reference standard defined the reported activity.

The report names “RS 03.” The receiving system contains two materials with that label. One is an early research lot. The other is a qualified reference material prepared later, with a different storage history. The result still has a number. The number still has decimals. A development team can spend weeks investigating a supposed assay discrepancy that began with a collision between two names.

An API contract should make that collision difficult. It should require a globally unique identifier or a resolvable local identifier, the identifier authority, the object type, the parent relationship, the method version, the unit, the transformation applied, and the person or system that asserted the result. A receiving organization can map those fields into its own data model while retaining the source identity and the mapping decision.

The contract also needs to say what happens when a field is unknown. Empty is different from not applicable. Unreported is different from never measured. “Same as previous batch” is not a value that a machine can safely inherit when the previous batch has three possible parents.

This sounds fussy until a regulatory reviewer asks why a release trend changed after a site transfer. The missing distinction then becomes a paragraph in a response package, followed by a reconstruction exercise involving archived emails.

Discovery data changes shape in manufacturing

Discovery systems tend to preserve possibility. Manufacturing systems preserve repeatability. Analytical development sits between them, translating a promising observation into a controlled measurement that can survive scale, site, and time.

RNA therapeutics make that translation especially demanding because the product story crosses sequence design, expression behavior, delivery chemistry, degradation, impurity profiles, and process conditions. A construct identifier alone cannot explain why two batches behave differently. The system must connect the construct to the formulation, the formulation to the process version, the process version to the batch, and the batch to the sample tested.

That graph should also connect study artifacts. A stability pull is a sample event. A chromatogram is an observation. A report is an interpretation of observations made under a declared method version. A deviation is an event that may alter how later results should be read. A regulatory table is a view over those objects, with its own generation timestamp and transformation record.

Calling the storage layer a knowledge graph does not create those relationships. I have watched teams put a graph label on a warehouse while the most important edge remained a human memory: this vial probably came from that batch. The warehouse held columns. The meaning lived in somebody’s retirement plan.

Federated systems can keep their borders

A shared semantic layer does not require Abogen and Novartis to replace their systems. It requires them to agree on the things that must remain recognizable across the boundary.

The originating laboratory can retain its electronic notebook, sequence repository, assay environment, and local workflow. The development organization can retain its study management system, quality platform, manufacturing data environment, and submission tools. A federated registry can resolve shared identifiers and expose only the attributes permitted by the collaboration agreement.

Security belongs inside the contract. A partner may receive a construct identity, provenance chain, and assay result while remaining unable to retrieve a protected sequence annotation or an internal project name. Attribute based access can govern the fields, while signed assertions record who supplied them and when. The API should return provenance with the data rather than offering provenance as an optional attachment that disappears during export.

The practical question is whether a recipient can answer five minutes later, without calling the originator, what this object is, where it came from, which transformation produced its value, and whether the value remains valid for the decision at hand.

That is a user experience decision as much as an architecture decision. Scientists need a readable explanation of identity and lineage, not a screen full of URI fragments. Engineers need machine resolvability. Quality teams need controlled change history. The interface has to serve all three without turning the human reader into a parser.

Broken meaning distorts development

Suppose an early study compares two formulations using samples whose container conditions were recorded under local labels. One laboratory maps “ambient hold” to a defined temperature range. Another maps the same phrase to a room temperature observation with no controlled duration. The resulting stability comparison favors a formulation that received gentler handling.

The development decision may be perfectly reasonable given the displayed data. The data still misleads because the terms were never made equivalent.

A second failure appears during scale up. The manufacturing team receives a process description and an analytical method, yet the critical sample preparation step refers to a concentration calculated from an intermediate volume. The originating laboratory had corrected for actual fill volume. The receiving site uses nominal volume. The method transfer produces a potency shift, and the team starts adjusting process parameters to repair an assay discrepancy that the data model could have exposed before the first run.

The regulatory narrative suffers differently. A submission may present a chain from candidate selection to clinical material, but the chain weakens when the identifiers change at each organizational boundary and the mappings live in manual reconciliation files. Reviewers do not need a grand theory of semantic systems. They need confidence that the sample tested in one table is the sample described in another, and that the reported result can be traced to a controlled method and a known batch.

“FAIR” on a slide does not supply that confidence. Ownership does.

The small agreement that changes everything

Teams stall because identity work feels like delay. Scientists want the next experiment. Engineers want a stable interface. Quality wants definitions approved before data moves. Each request sounds reasonable, especially when the collaboration has a milestone attached to it and everyone is tired of meetings about nouns.

A smaller agreement can break the deadlock. Pick one handoff, such as the transfer of an analytical method from discovery into development. Define the objects that cross it. Give each object an identifier policy. Specify the minimum provenance. Version the API contract. Test the exchange against a real exception, such as a reprocessed sample, a superseded method, or a batch split across laboratories.

The test should ask whether a receiving scientist can distinguish the original observation from the transformed result and the approved conclusion. If the answer requires opening three systems and asking the originating scientist to remember why a file was renamed, the handoff has already failed.

The next construct will expose a different gap. Good. The gaps should become visible while they are still cheap.

ABO2203 may advance through development, and additional programs may emerge from the option structure around Abogen’s RNA platform. Each future program will carry its own scientific uncertainties. Nobody needs to add identity uncertainty as a bonus feature.

The RNA Development Identity Authority remains imaginary, which is precisely why organizations keep paying for its absence through duplicated assays, delayed transfers, disputed batches, and regulatory explanations assembled after the fact. Build the shared meaning first, then let the systems keep their borders.

If this handoff problem is sitting on your desk, hello@example.com is where the conversation can begin.