Before writing new guidance for fermentation dossiers, the Food Standards Agency did something unusually specific: it went back through the applications it already had and worked out where they kept getting stuck. That scoping exercise produced four recurring problem areas — bioinformatics, protein digestibility related to nutritional quality, allergenicity, and clarity on the identity and composition of the product.
Expert elicitation events followed in 2025. The resulting document, Supplementary Guidance for Novel Foods Produced by Precision or Biomass Fermentation, was published on 20 March 2026 in FSA Research and Evidence under DOI 10.46756/001c.158776, having been reviewed by the Advisory Committee on Novel Foods and Processes at its 175th meeting in February.
The interesting part is what happened next. The committee’s minutes from the meetings either side of publication are public, and the live dossiers in them are failing on precisely those four things.
What the guidance actually requires
This is not a policy statement. It is a technical specification, and most of the reporting on it has not gone past the abstract. The numbers below are the ones that will decide whether a dossier moves.
Protein digestibility: the 12% line
The guidance introduces a proportionate, tiered test for whether a digestibility assay is needed at all, and it attaches illustrative numbers to it:
- Low-inclusion techno-functional uses are considered to be around 5% or less inclusion in the final product.
- Uses as a source of protein are considered to be around 12% or more.
The FSA is explicit that these are illustrative benchmarks and that the decision is driven by the ingredient’s actual role and the resulting exposure, not by the percentage alone. But it is the first published number of its kind we have seen from a GB regulator, and it converts an argument into a threshold.
Above that line, the guidance is unambiguous about method:
| Requirement | What the guidance says |
|---|---|
| Primary score | DIAAS (Digestible Indispensable Amino Acid Score) |
| Supportive metrics | PDCAAS, Digestible Indispensable Amino Acid Ratio, amino acid profiles, nitrogen data |
| Default method | INFOGEST static protocol, in vitro |
| In vivo | Pig models reserved for specific unresolved questions or to validate in vitro methods; small rodent studies discouraged |
| Replicates | At least 3 technical replicates |
| In-run references | One high-digestibility reference (milk/casein or skim milk powder), ideally one moderate (soy), digested alongside the test item |
| Literature values | Cannot substitute for in-run references |
| Comparator | The protein the ingredient is intended to replace (casein for dairy analogues) |
That in-run reference rule is the sharpest edge in the document. The stated reason is that “the assay conditions and analytics drive the outcome”, so a DIAAS figure from a paper cannot be set against a DIAAS figure from your own run. Anyone whose dossier cites a published comparator number rather than one generated alongside the test material has a rewrite ahead of them.
Practical consequence: if your ingredient is a protein source rather than a texturiser, budget for INFOGEST work with in-run casein and soy references and LC-MS/MS amino acid analysis on the same workflow, and do it before submission rather than in response to a question. This is directly connected to the basis problem we described in crude protein versus true protein in microbial biomass: DIAAS is what a regulator asks for when nitrogen-based protein figures stop being persuasive.
Identity: whole genome sequencing, in detail
| Organism type | What is expected |
|---|---|
| Bacteria and yeasts | A complete closed whole genome sequence is standard; justification required if not supplied. For bacteria, plasmid and non-nuclear DNA must be accounted for |
| Plasmids | State whether the production strain contains them; give size range estimates and sequencing data, with methods adjusted for small plasmids and dual long-and-short-read approaches for large ones |
| Fungi | Ploidy level should be defined; sequencing of non-nuclear DNA not normally required |
| Filamentous fungi | WGS still recommended, but recognised as not always feasible — provide justification |
| Identity confirmation | Whole-genome approaches, ideally with at least two independent methods |
| Phylogenetic tree | Provide alignments plus a narrative on construction, sister groups, choice of reference strains, rooting and method. Maximum Likelihood Estimation recommended |
| GMOs | Where modification occurred, how it was achieved, the modified sequence, and whether it is integrated or plasmid-borne |
| Unintended integration | Evidence on whether additional integration events occurred, and a complete analysis of likely impact if so |
There is also a QPS trap worth naming. If a production organism has qualified presumption of safety status, applicants must say so and say whether it was granted for the organism as a whole or only as a production organism — and the guidance states that the genome-based evidence “should still be provided regardless of the QPS status of the organism.” QPS reduces the argument. It does not remove the sequencing.
Production process and residual host material
- Kill-step validation with a sampling plan, heat treatment conditions linked to microbial reduction, and HACCP documentation — with batch-level evidence, typically from 5 batches, particularly for biomass products.
- Shotgun proteomics recommended to identify residual proteins derived from the production organism, after initial quantification.
- Residual DNA checks scaled to product type: reduced risk for highly purified products but verification still required; more stringent checks for biomass products.
- Genetic drift: applicants should state how many fermentation cycles the same inocula is used for before restarting with new inocula. Notably, the guidance says whole genome sequencing “would not be expected to be used to evaluate genetic drift” — a rare instance of a regulator telling applicants what not to spend money on.
Scope — read this before assuming it applies to you
Organisms in scope are selected non-GM microorganisms and category 1 and 2 GM microorganisms, where no GM DNA content is present in the final product, covering bacteria, yeasts and filamentous fungi.
Out of scope: microalgae, plants, animal cells, category 3 and 4 GM organisms, and precision bred organisms. The FSA notes the list is not exhaustive and that some points may still be relevant to excluded types.
Practical consequence: gas fermentation using non-listed organisms, microalgal proteins, and anything where modified DNA persists in the final product sit outside this document. If that is your product, the supplementary guidance is informative but not addressed to you.
The same four gaps, in live dossiers
Here is where the committee minutes become the evidence. The table below is assembled from the published minutes of the 175th meeting (11 and 12 February 2026) and the 176th (23 April 2026). Every entry is the committee’s own characterisation.
A note on the quotations. GOV.UK renders these minutes with acronym expansions inserted inline, so the source text reads “BLG (Beta-Lactoglobulin)” and “GMO (Genetically Modified Organism)”. We have removed those inserted expansions from quoted passages and changed nothing else. Any quoted string here is otherwise verbatim.
| Application | Reviews to date | Gaps the committee identified |
|---|---|---|
| Fermotein RP1215 (Rhizomucor pusillus biomass) | Fourth and final review, February 2026; CAD agreed | Production process, composition, specification and ADME at the third review. At the fourth: not appropriate to call it a “high protein source” as it is not highly digestible; margin of safety for toddlers, children and adolescents smaller than required |
| Beta-lactoglobulin RP1571 (precision fermentation) | Third review, April 2026; still open | Protein content is predominantly BLG “along with fungal proteins derived from the genetically modified organism”; identity and specification to be updated to reflect this; inorganic arsenic traced to the filtration aid; mycotoxin method validation not available; elevated acute allergen exposure versus milk powder |
| Solein RP1326 | Second review; returned for more data | Identity, production process, specification, proposed use levels, nutrition, toxicological information |
| Fusarium str. flavolapis RP1637 | Second review; returned for more data | Identity, production process, specification, proposed uses, nutrition, toxicological information |
| Human-equivalent lactoferrin alpha RP2296 | First review, April 2026 | Protein structure, iron binding and glycosylation; iron exposure and nutrient interactions; immunotoxicology; allergenicity. Hazard identification deferred to the next meeting on volume and complexity grounds |
Map that against the four scoping areas and the fit is close to exact.
Identity and composition. The beta-lactoglobulin finding is the cleanest illustration in the whole record: analysis requested by the committee confirmed that the protein content consists predominantly of BLG along with fungal proteins from the production organism. That is not a contaminant story — it is the specification being wrong about what the product is. It is exactly what the guidance’s shotgun-proteomics recommendation exists to catch, and it is why “purified protein” is a claim that now has to be demonstrated rather than asserted.
Protein digestibility. Fermotein’s outcome is the clearest possible statement of why DIAAS matters commercially. The committee concluded the novel food is safe for adults under the proposed uses, but recorded that the term “high protein source” is not appropriate for this product because it is not highly digestible, and identified a potential nutritional disadvantage if used as the primary protein source in the diet. A product can pass safety and still lose the claim that makes it saleable. We covered The Protein Brewery’s separate EU approval and closed US notice; this is the third jurisdictional data point on the same ingredient, and the one that touches the label.
Allergenicity. On beta-lactoglobulin, the committee noted an elevated risk of serious reactions because acute exposure from the novel food is higher than from consuming the same amount of milk powder in the same food categories, and flagged the product as a potential “hidden food allergen” in categories where milk is not normally used. It referred the question to risk managers. That is the dossier-level counterpart to the labelling divergence between FSANZ and the UK we wrote about last week — and it shows the science committee reaching a concern before the labelling rules have settled.
Bioinformatics. This one is visible mostly by absence: the guidance devotes more space to WGS than to any other topic, which is consistent with a scoping exercise that found it recurrently problematic, but the published minutes do not itemise WGS failures by applicant. We are not going to claim evidence we do not have.
Britain is still assessing against the 2016 guidance
One decision from the 175th meeting deserves more attention than it has received. The committee reviewed EFSA’s updated 2024 novel food guidance and considered whether it should form the basis for full novel food assessments in Great Britain.
It decided against. Members agreed that the 2016 guidance should continue as the basis for GB assessment, noting that the primary legislation is unchanged between GB and the EU, and that gaps in the 2016 guidance could be addressed through existing FSA supplementary guidance and targeted requests for additional data. The decision is to be kept under review “in light of the EU reset and work to agree a SPS agreement with the EU”. An intersessional review of the compositional data and specifications sections was actioned.
Practical consequence: a dossier written to EFSA’s 2024 guidance is not automatically aligned to GB expectations, and vice versa. The base document in Great Britain is the 2016 guidance plus supplementary layers — of which the March 2026 fermentation document is now one. Anyone sequencing filings across the gap between EU and US clearance rates should treat GB as a third specification, not a copy of either.
Alongside the guidance, the committee finalised a Committee Advice Document template for precision fermented novel foods, described as an internal document to provide structure for assessments and address common issues identified during the Innovation Research Programme. That is a quieter but arguably more consequential change: it standardises how these products are assessed internally, which tends to shorten review cycles once applicants learn its shape.
The counter-argument
The obvious objection is that this is a good-news story dressed as a problem. A regulator that studies its own bottlenecks, publishes what it found, names its preferred assay, and gives an illustrative inclusion threshold is doing more for applicants than one that issues questions case by case. The FSA has also stood up a Precision Fermentation Business Support Service. Compared with a process where every applicant discovers the digestibility requirement by being asked for it at the second review, published guidance is a straightforward improvement.
That is fair. The counter to the counter is cost and timing. DIAAS by INFOGEST with in-run references, complete closed genome sequencing with plasmid characterisation, shotgun proteomics on the final product and five-batch kill-step validation is a substantial analytical programme, and the guidance arrived in March 2026 — after several of the dossiers now in the queue were written. The companies in the table above are absorbing the cost of a standard that was documented after they filed.
It is also worth noting the guidance’s own limit, stated plainly in the document: following it “does not necessarily guarantee that a product will be authorised”, and authorisation is a matter for Ministers.
What we could not establish
- How many applications the scoping exercise reviewed, or which ones. The scoping report is public but we did not fetch and read it in full for this piece, and we are not going to characterise its sample from a summary.
- The identity of the applicants behind RP1571 and RP2296. The minutes use reference numbers; we did not find a public register entry tying these to companies and are not going to guess.
- Whether the arsenic finding in RP1571 exceeds any regulatory limit. The minutes record the source as the filtration aid and the committee’s concern under the ALARA principle. They do not state a concentration or a limit.
- Whether Fermotein’s UK outcome permits sale. The 176th minutes state the CAD was agreed and that “the next steps for this application are being reviewed”. A CAD is committee advice, not an authorisation.
- Any WGS-specific deficiency by applicant. The guidance implies these are common; the published minutes do not itemise them.
- The minutes carry the standing note that they are subject to confirmation by the Committee. We have quoted them as published.
One small thing that suggests these documents are read less than they should be: the 175th minutes head the Fermotein item “Fermotein RP12151”, while every other reference in both sets of minutes says RP1215. It is a typo, and it has survived publication.
What to watch
- Whether beta-lactoglobulin RP1571 returns for a fourth review and whether the arsenic and identity points are resolved or escalated to risk managers. The committee has already referred the allergen question upward.
- Whether human-equivalent lactoferrin alpha RP2296 completes hazard identification. The committee deferred it once already on complexity grounds. A second deferral would say something about the assessment capacity available for structurally complex proteins.
- Whether any applicant publishes a DIAAS figure generated under INFOGEST with in-run references. The moment one does, it becomes the number every other fermentation protein is compared against — and on current evidence, very few will be pleased with the comparison.
- The intersessional review of the compositional data and specifications sections of EFSA’s 2024 guidance. If GB moves to the 2024 basis, dossier requirements shift again.
The wider point is that the constraint on fermentation-derived protein in Great Britain has stopped being conceptual. Nobody in these minutes is arguing about whether a microbial protein can be a food. They are arguing about ploidy, plasmid size ranges, filtration aids, in-run reference proteins and whether toddlers have enough margin. That is what a regulated category looks like from the inside, and it is considerably more expensive than the category expected.