Can Anyone Tell How Your New Protein Was Checked for Allergens?

Dr. Bryan Quoc Le · Mendocino Food Consulting · November 2026

For most GRAS notices on new proteins, the answer is no. FDA's letter says the protein is unlikely to be allergenic, but it almost never says how anyone reached that conclusion. I went through 83 notices to see what the public record actually shows, and what a notifier can do about it.

Why this matters

Every new food protein raises the same question: could it trigger an allergic reaction, or cross-react with an allergy someone already has? That goes for an animal-free whey protein, a mycoprotein, or a new plant protein isolate.

There's a well-established way to answer it. You compare the protein's sequence against a database of known allergens, check how much of the production organism's own protein is left in the ingredient, see whether the protein survives digestion, and test with blood from allergic people when the source is a known allergen. The method comes from FAO/WHO (2001) and Codex (2003).

But for anyone outside FDA (an investor, a customer, a retailer, a competitor, a food allergy group), what matters is whether they can see that work. For most notices, the only document that's easy to find is FDA's response letter.

What I did

I pulled every GRAS notice in FDA's inventory for proteins, precision-fermentation proteins and biomass ingredients: 138 notices. For 83 of them I could get FDA's letter, the notifier's dossier, or both. The other 55 are older notices whose letters sit in FDA's web archive.

Each notice was coded on 13 points: what the protein is, whether a sequence screen was done and described, which database and thresholds were used, host-cell proteins, digestion and heat testing, IgE or clinical testing, labeling, and what FDA's letter says. Two AI coders worked independently, a third broke ties, and I gave the results a brief review.

Then I gave each notice a simple transparency score out of five, one point each for a described screen, a named database, stated thresholds, quantified host-cell proteins and a digestion test. It doesn't grade whether the science is good. It counts whether a reader could check it.

The full methods and limitations are in the technical report. To see how your own notice stacks up, use the free Allergenicity Evidence Checklist.

What I found

1. FDA's letters tell you the conclusion, not the method. For 48 notices, FDA's letter is the only document I could get. Not one of them names the allergen database or the screening thresholds. Most "no questions" letters do talk about allergenicity, but it's usually FDA repeating the notifier's conclusion plus a standard paragraph about allergen labeling.

This isn't FDA hiding anything. The dossiers I could read are often thorough. The letter just isn't written to carry the method, and for older notices the letter is all that's posted.

2. In the dossiers, ingredient type made the biggest difference. Across the 35 notices with a dossier, the average score was 2.5 out of 5. But that average hides two very different groups.

Precision-fermentation proteins and proteins identical to a known allergen (animal-free β-lactoglobulin, casein, ovalbumin) were documented well. Their notifiers screened the yeast or fungal proteins left in the ingredient, measured them, and usually tested digestion. Conventional protein isolates and concentrates (rice, hemp, fava, almond, goat whey and similar) mostly scored zero. They rested on the history of eating the source food, which is a reasonable argument, but they rarely made it explicit, or asked whether concentrating or hydrolyzing the protein changes anything.

3. Almost everyone uses the same database and the same 2003 rule.

All 83 notices with an FDA letter or a dossier. A version or date matters because AllergenOnline is updated every year; without it, nobody can repeat the screen.

No notice built its assessment on anything newer than the Codex (2003) and FAO/WHO (2001) guidance, even though EFSA questioned the 35% rule in 2022.

4. Testing beyond the screen is rare. About a quarter of notices (20 of 83) reported a digestion test. Five reported heat or thermal testing. Only one reported any human immunological testing.

5. Labeling often goes unmentioned. 55% of notices didn't say what the label would declare. Proteins identical to a listed allergen did better: 71% committed to declaring it, for example as "milk" or "egg".

What to put in your allergenicity section

None of this requires new testing. Most of it is reporting what you already did, in one place.

• Say what the protein is. Its sequence identity, its source, and whether it's identical or similar to a known allergen.
• Describe the screen in two sentences. What you compared (the protein, the host-cell proteins, or the whole predicted proteome) against which database, and which version or search date.
• State the thresholds. More than 35% identity over 80 amino acids, plus an 8-amino-acid exact match. It costs nothing to add.
• Quantify what else is in there. Host-cell or co-isolated proteins as a percent of total protein, and how they were screened.
• Report digestion and heat data. Even a protein digestibility assay you ran for nutrition counts. Say what it showed.
• Make the history-of-use argument explicit if that's your case, and address whether your process could change allergen levels compared with the source food.
• Say what the label will declare, or why no allergen declaration is needed.
• Put it all on one page. A short allergenicity summary that a reviewer, or anyone reading your dossier later, can find without digging through appendices.

I'd also like to see FDA add one sentence to its letters naming the methods it reviewed. It would turn the most-read document in the record into one that actually shows the work.

About the author

I'm Dr. Bryan Quoc Le, a food scientist and the founder of Mendocino Food Consulting. I've spent more than ten years in food product development, and I'm the author of 150 Food Science Questions Answered. I've worked with companies from early-stage startups to Blue Diamond Growers and The Good Food Institute.

Working on a GRAS notice for a new protein? Send me a short description of your ingredient and where your allergenicity section stands at bryan@mendocinofoodconsulting.com, and I'll tell you what a reviewer is likely to ask. That's the core of my Second Opinion Call.