The fastest vaccine development in history also revealed, in stark detail, the infrastructure failures that structured biological design data would have reduced. This is the case study the field needs to examine honestly.
The COVID-19 pandemic mRNA vaccine scale-up was the most ambitious biological manufacturing operation in human history. It was also an involuntary, real-time demonstration of what happens when complex biological systems are transferred between facilities using documentation designed for a different era. This article examines the scale-up in detail: the technical challenges of LNP manufacturing transfer, the documented failures at Emergent BioSolutions and elsewhere, the regulatory response, and the specific points at which structured biological design data — SBOL-compliant process specifications, formal characterisation assertions, and machine-readable CMC documentation — would have reduced the failures that occurred. It also examines what the mRNA platform has taught the field about the relationship between design precision and manufacturing consistency, and why the lessons are still not being systematically applied.
Keywords
1. The Triumph and Its Shadow
The first article in this series argued that biology has built discovery infrastructure but not design infrastructure — and that the consequences are visible, documented, and preventable. This article examines the most instructive case study available: the COVID-19 mRNA vaccine scale-up, which compressed years of normal manufacturing development into months and made the documentation failures of the biological design system impossible to ignore.
On 8 November 2020, Pfizer and BioNTech announced that their mRNA vaccine candidate had demonstrated 90% efficacy in Phase III trials. On 11 December 2020, the FDA granted Emergency Use Authorisation. From viral sequence publication to authorised vaccine had taken approximately eleven months.
The speed was a genuine scientific and organisational achievement. It drew on decades of foundational mRNA research, the accumulated institutional knowledge of clinical trial infrastructure, and the willingness of regulatory agencies to move at a pace they had never previously attempted. It also drew on the decision, made early in the pandemic, to begin manufacturing scale-up before Phase III results were available — a bet that cost billions and was vindicated.
What followed is less well documented, but more instructive. The process of scaling from clinical trial supply to billions of doses revealed infrastructure failures that the speed of the initial development had concealed.
This article examines those failures not to diminish the achievement, but because understanding them precisely is the only way to prevent their recurrence. The mRNA platform is not going away. LNP-based therapeutics are the most active area of drug development in the world right now. The manufacturing and documentation failures of the COVID scale-up will recur in increasingly consequential contexts unless the field addresses them structurally.
2. The LNP Manufacturing Challenge
2.1 Why LNPs Are Hard to Transfer
Lipid nanoparticles are thermodynamically metastable structures. They do not exist in nature and they do not spontaneously form under most conditions. They are assembled through a specific manufacturing process — typically microfluidic mixing of lipids in an organic solvent with nucleic acid in an aqueous buffer — and their properties are acutely sensitive to the conditions of that assembly process.
The same four-component lipid formulation — ionisable lipid, phospholipid, cholesterol, PEG-lipid — will produce particles of different sizes, different encapsulation efficiencies, and different transfection properties depending on the flow rates used during mixing, the temperature of the inlet streams, the concentration of the lipid solution, the N/P ratio of lipid to nucleic acid, and the downstream buffer exchange conditions. These dependencies are partially understood theoretically and partially characterised empirically for specific formulations, but the relationship between process parameters and product properties is not fully predictable from first principles.
This means that manufacturing transfer for an LNP system is not a matter of sending a recipe and having a competent facility follow it. It is a matter of transferring sufficient process understanding — including the empirical knowledge about which parameters matter most, which interactions between parameters produce critical quality attributes, and which deviations are consequential — for the receiving facility to reproduce not just the procedure but the understanding that makes the procedure work.
Manufacturing transfer for an LNP is not a matter of sending a recipe. It is a matter of transferring sufficient process understanding for the receiving facility to reproduce not just the procedure but the knowledge that makes it work.
2.2 The Documentation Gap
The documentation packages used for COVID vaccine manufacturing transfers were, by the standards of pharmaceutical CMC documentation, unusually detailed. The urgency of the situation had driven developers to document more thoroughly than is typical in early-phase development. But they were still, fundamentally, prose-based PDF documents — descriptions of a process rather than formal specifications of it.
The distinction matters at the level of critical process parameters. A process description might state that mixing should occur ‘at a flow rate ratio of approximately 3:1 aqueous to organic.’ A formal process specification would state that the aqueous-to-organic flow rate ratio should be 3.0 ± 0.1, measured by the specific flow controllers installed in the manufacturing system, with a defined response protocol if the ratio deviates outside this range. The difference between these two specifications is not merely precision — it is the difference between a document that a facility can implement consistently and one that requires interpretation.
Interpretation introduces variability. Variability produces batches with properties outside the specified range. Batches outside specification are either rejected or, if the out-of-specification investigation is inadequate, released with compromised quality. The COVID scale-up produced all three outcomes.
3. What the Documentary Record Shows
3.1 The Emergent BioSolutions Case
The Emergent BioSolutions facility in Bayview, Baltimore, was contracted by both AstraZeneca and Johnson & Johnson to manufacture COVID-19 vaccine drug substance. In March 2021, it was reported that approximately 15 million doses of Johnson & Johnson vaccine had been contaminated with AstraZeneca material — a cross-contamination event that led to the immediate suspension of manufacturing at the facility and an FDA investigation.
The FDA’s inspection, conducted in April 2021 and documented in Form 483 inspectional observations, identified multiple deficiencies. These included inadequate procedures for the prevention of cross-contamination between vaccine programmes, deficient environmental monitoring, and — critically for our purposes — insufficient documentation of critical process parameters and their acceptable ranges. The inspection report noted that the facility’s procedures did not adequately define the process controls necessary to ensure consistent manufacturing, and that deviations from expected process conditions were not being detected, documented, or investigated systematically.
This is a documentation failure of exactly the kind that structured process specifications are designed to reduce. Had the critical process parameters for each vaccine programme been formally specified — as SBOL-compliant data objects with defined acceptable ranges and mandatory deviation reporting triggers — the monitoring and documentation gaps that allowed the contamination event to develop undetected would have been far harder to sustain.
3.2 The Broader Pattern
Emergent BioSolutions was the most visible failure, but it was not isolated. Industry reporting throughout 2021 and 2022 pointed to elevated batch rejection rates at several newly commissioned facilities, with process deviations at new LNP manufacturing sites frequently traced to ambiguities or gaps in the manufacturing documentation transferred from the originating facility.
The specific failure modes recurred across facilities and vaccine programmes: inconsistent particle size distributions attributed to variability in microfluidic chip installation and qualification; encapsulation efficiency below specification attributed to variation in lipid stock preparation and N/P ratio calculation; and post-formulation stability failures attributed to inconsistencies in buffer exchange conditions and storage temperature during transfer.
In each case, the originating facility had the process knowledge to prevent the failure. The knowledge had not been transferred — not because the transfer team was incompetent, but because the documentation format used did not have the structure to carry it faithfully.
4. What Structured Design Data Would Have Changed
4.1 At the Point of Transfer
A manufacturing transfer package built on the Nanomedicine Design Stack would make L1 — Manufacturing & Specification explicitly traceable to the L2 Nanomaterial Architecture it is intended to realise. Critical LNP process parameters could be encoded as formal SBOL3-compatible data with defined Measure values, acceptable ranges and links to characterisation assertions showing how those parameters affect critical quality attributes. The transfer package would therefore carry not only a recipe, but the structured relationship between the material design, its manufacturing conditions and the evidence supporting them.
The receiving facility’s quality management system would have been able to ingest this data computationally, map the specified parameters to its own equipment configuration, identify the parameters for which its equipment did not match the specification, and flag these for process development investigation before manufacturing began — rather than discovering them as batch failures after manufacturing had started.
This is not speculative. It is the same computational comparison that SBOL-enabled design tools already perform for synthetic biology circuits when a design is transferred between laboratories. The technical infrastructure exists. The application to pharmaceutical manufacturing transfer is an extension of existing capability, not a new development.
4.2 At the Regulatory Level
The FDA’s post-pandemic review of COVID vaccine manufacturing failures has informed its current thinking on process validation and CMC documentation requirements for complex biological products. The agency’s emerging guidance on process analytical technology and real-time release testing for advanced therapy medicinal products reflects a recognition that the documentation practices of the pre-pandemic era are insufficient for the manufacturing complexity of the current pipeline.
Structured design data would have supported this regulatory oversight directly. An agency that could ingest machine-readable process specifications from multiple manufacturing sites, compare them computationally against the approved design, and identify deviations automatically would have been able to identify the Emergent BioSolutions documentation deficiencies without a physical inspection — or at minimum, would have been directed by computational anomaly detection to conduct a targeted inspection earlier in the contamination sequence.
5. The mRNA Platform Going Forward
5.1 The Pipeline That Is Coming
The mRNA vaccine platform that was stress-tested by COVID is now being applied to influenza, RSV, HIV, malaria, and personalised cancer vaccines. LNP-based mRNA therapeutics are in clinical development for a range of rare diseases, cardiovascular conditions, and hepatic disorders. The manufacturing complexity of these second-generation applications is, in most cases, substantially greater than the COVID vaccines — more complex lipid formulations, more sensitive mRNA cargo, more demanding cold chain requirements, and more stringent efficacy thresholds that leave less tolerance for batch-to-batch variability.
The documentation failures that produced batch losses and contamination events during the COVID scale-up will recur in these programmes unless the field implements structured design documentation as standard practice. The consequences will be more severe: in a therapeutic context, a batch failure is not just a supply problem — it is a patient access problem, and in rare diseases where alternative treatments do not exist, it may be a patient safety problem.
5.2 The Industry Response
There are grounds for qualified optimism. Several of the larger mRNA developers — Moderna, BioNTech, and Arcturus Therapeutics — have invested significantly in manufacturing informatics infrastructure since 2021, and public statements from their manufacturing leadership suggest that structured process specification is a priority. The NIIMBL manufacturing innovation institute in the US has funded research into digital manufacturing standards for biologics that is directly relevant to LNP manufacture.
What is missing is a common standard that allows the process knowledge developed at one organisation to be transferred to another in a format both can computationally process. Individual investments in internal data infrastructure are valuable but insufficient — the transfer problem is an inter-organisational problem, and it requires an inter-organisational solution. SBOL, extended to cover pharmaceutical manufacturing parameters, is the most credible candidate for that solution currently available.
6. Conclusion: Learning the Right Lesson
The COVID mRNA vaccine scale-up is routinely cited as evidence of what the life sciences can achieve when the barriers to speed are removed. That is true and worth celebrating.
It should also be cited as evidence of what happens when a novel biological platform is transferred at global scale using documentation infrastructure that was not designed for the task. The contaminated batches, the process deviations, the batch rejections — these were not the inevitable cost of moving fast. They were the preventable cost of moving fast with inadequate infrastructure.
The lesson the field needs to learn is not that speed requires sacrifice of quality. It is that the infrastructure investment required to maintain quality at speed was not made before it was needed — and that making it now, for the programmes that are currently in development, is the only responsible path forward.
We did not hit a scientific limit during the COVID scale-up. We hit an infrastructure limit. The science was ready. The documentation system was not.
The cost of not learning that lesson is already visible. It is measurable in the batch losses of 2021. It will be measurable again, in more consequential terms, in the therapeutic failures of programmes that are currently in clinical development if the field does not act.
References
- U.S. Food and Drug Administration (2021). Form 483 Inspectional Observations: Emergent BioSolutions, Bayview, Baltimore, April 2021.
- Kulkarni, J. et al. (2021). The current landscape of nucleic acid therapeutics. Nature Nanotechnology, 16, 630–643.
- Schoenmaker, L. et al. (2021). mRNA-lipid nanoparticle COVID-19 vaccines: structure and stability. International Journal of Pharmaceutics, 601, 120586.
- Hou, X. et al. (2021). Lipid nanoparticles for mRNA delivery. Nature Reviews Materials, 6, 1078–1094.
- Pardi, N. et al. (2018). mRNA vaccines — a new era in vaccinology. Nature Reviews Drug Discovery, 17, 261–279.