The reproducibility challenges in nanomedicine are routinely framed as a scientific limitation. Increasingly, both academic literature and industrial experience suggest that this framing is incomplete. In most cases, the underlying problem is not the science. It is the way designs are described — and that is a problem the field has the tools to fix right now.

Ashley Morgan · April 2026 · 8 min read

Picture the following scene. A research group in Helsinki has spent three years developing a lipid nanoparticle formulation that delivers siRNA to hepatocytes with remarkable efficiency. They publish. The methods section is careful, the supplementary data thorough. A group in Rotterdam reads the paper and attempts to reproduce the formulation. They follow the protocol closely. They obtain something different — particle size distribution shifts, encapsulation efficiency drops, in vivo behaviour diverges.

A short, polite argument about whose equipment is calibrated correctly plays out across two journal issues and eighteen months. Both groups are competent. Neither is careless. The results do not align.

Situations of this kind are well documented across nanoparticle systems, particularly during scale-up and cross-site transfer. Small variations in mixing kinetics, solvent exchange rates, lipid phase behaviour, or temperature control can produce measurable changes in particle size, surface charge, and payload encapsulation — even where nominal compositions are identical. In lipid nanoparticle formulations specifically, molar ratios do not simply define composition. They interact with ionisable lipid pKa, PEG-lipid desorption kinetics, and microfluidic process parameters — flow rate ratio, total flow rate, aqueous-to-organic mixing conditions — in ways that collectively determine particle size distribution, stability, and biological behaviour. Even nominally identical formulations can diverge due to differences in microfluidic mixing regimes, such as variations in Reynolds number or solvent polarity gradients during nanoprecipitation. A small ambiguity at the point of description propagates into a significant divergence in outcome.

What appears to be a shared formulation is, in practice, an under-specified design. The problem is not that the science fails. It is that the description permits multiple valid interpretations.

It would be incorrect to suggest that all irreproducibility in nanomedicine arises from descriptive limitations. Biological variability and process sensitivity remain significant contributors. However, where ambiguity in design representation exists, it compounds these challenges and makes systematic improvement considerably more difficult.

A nanoparticle formulation that behaves differently across manufacturing sites is not a research inconvenience. It is a patient safety concern.

The description problem

Mature engineering disciplines confronted this challenge decades ago. As systems increased in complexity, prose descriptions proved insufficient. In response, formal, machine-readable design representations emerged — CAD files in mechanical engineering, SPICE netlists in electronics, standardised schematics in architecture. These formats are not merely convenient. They are foundational. They constrain ambiguity, enable verification, support version control, and allow designs to be transferred between organisations without loss of meaning.

Nanomedicine, by contrast, still relies heavily on narrative description. A nanoparticle therapeutic is communicated through a combination of methods sections, figures, and supplementary tables — formats that were never designed to capture high-dimensional design spaces or tightly coupled process dependencies. They are not computationally interoperable, not version-controlled in any meaningful sense, and not capable of enforcing internal consistency.

The gap between what can be built and what can be reliably communicated is widening every year. We can design and deploy complex nanoparticle delivery systems at global scale. We can characterise biological interactions at molecular resolution. We can manufacture at industrial volumes with remarkable precision. And yet, in many cases, the primary method for transferring a nanoparticle design between research groups — or between a development team and a contract manufacturer — remains a Word document, or a phone call to the corresponding author asking what they actually did, because the paper does not quite say.

This is 1995-era information infrastructure carrying 2026-era scientific complexity. The consequences are cumulative: irreproducible results, delayed technology transfer, increased regulatory burden, and a machine learning potential that cannot be realised because the underlying data does not exist in a computable form.

The regulatory dimension

The regulatory consequences of this infrastructure gap are, if anything, more acute than the scientific ones. Chemistry, Manufacturing and Controls (CMC) documentation — the submission that sits at the heart of every regulatory filing with the MHRA, EMA, or FDA — requires that a therapeutic product’s composition, design, and manufacturing process be documented with sufficient precision to ensure consistency across batches, sites, and time.

Within ICH Q8 (Pharmaceutical Development) and Q10 (Pharmaceutical Quality Systems), this expectation is formalised through Critical Quality Attributes (CQAs) and Critical Process Parameters (CPPs). For nanoparticle systems, these attributes are highly sensitive to both formulation and process conditions. A prose-based CMC submission introduces ambiguity at precisely the point where regulators require determinism, particularly when defining design space and demonstrating process robustness.

Such submissions are often accepted today — but this reflects the absence of established alternatives, not the adequacy of the approach. Regulatory thinking is already evolving. Quality by Design (QbD) frameworks emphasise structured understanding of design space and process control. As expectations shift toward greater traceability and data integrity — aligned with ALCOA++ principles — the limitations of narrative documentation will become increasingly visible.

The organisations that adopt structured, machine-readable design representations early will find regulatory interactions more efficient. Those that do not will find themselves retrofitting clarity under pressure — during a regulatory review, at a late stage of development, when the cost of ambiguity is at its highest.

A language that already exists

The Synthetic Biology Open Language — SBOL, currently in its third major version — was developed by a community of researchers and engineers over more than a decade to solve exactly this class of problem in synthetic biology. It provides a formally specified, machine-readable framework for describing biological designs: not just sequences, but functional roles, component relationships, provenance, experimental context, and intended behaviour.

SBOL is not a database or a piece of software. It is a standard — in the same tradition as HTML, which nobody owns but which made the World Wide Web possible. Critically, adopting SBOL does not require openness. It requires interoperability. A company can use it internally, communicate with contract manufacturers, satisfy regulatory requirements, and participate in pre-competitive data sharing without publishing proprietary formulations. The standard is the language. What is written in that language remains the author’s to control.

SBOL3 is not a finished solution for nanomedicine. Its current adoption in this domain is limited, in part due to differences in abstraction and the absence of domain-specific ontologies for nanoparticle systems. Tooling remains uneven. But its underlying architecture is well suited to extension — and that extension is under active development.

The Nanomedicine Design Stack — a framework developed by Molecular Precision to describe the five layers at which design decisions are made in nanoparticle therapeutic development — maps directly onto SBOL3’s extensible data model. Layer 1, molecular composition, already has a working reference implementation: a structured SBOL3 encoding of an mRNA lipid nanoparticle formulation based on the publicly documented SM-102 four-component system. It demonstrates what becomes possible when a formulation is described in a format that is machine-readable, version-controlled, and computationally verifiable for internal consistency.

The field does not primarily have a science problem. In many cases, it has a description problem — and the tools to solve it already exist.

What needs to happen

The barriers to adoption are real. They are not, however, scientific. The tooling gaps are engineering problems. The incentive misalignments — researchers are rewarded for publications, not for rigorous data infrastructure — are institutional problems. The disciplinary fragmentation between the nanomedicine and SBOL communities is a coordination problem. None of them require new science to resolve.

A realistic path forward is incremental:

The organisations that invest early are not simply improving documentation. They are building a proprietary dataset that is internally consistent, computationally queryable, and capable of supporting machine learning workflows that competitors operating with fragmented documentation cannot match. This is not an argument from altruism. It is an argument from strategic positioning in a field where data quality will increasingly determine the pace of development.

The mRNA vaccine programmes demonstrated what is possible when a delivery platform reaches clinical maturity at speed. The next programmes — in oncology, rare disease, gene therapy, infectious disease — will operate in a world where regulatory expectations are higher, manufacturing complexity is greater, and the pressure to move fast is undiminished. The field that enters that world still describing its designs in narrative prose will find it considerably harder than it needs to be.

The infrastructure exists. The standard is ready. The remaining constraint is not technical feasibility, but adoption.

The question is no longer whether nanomedicine can be described with engineering-grade precision — but how long the field will continue to operate without it.