Popular Rationalism

James Lyons-Weiler, PhD

There is a question I have been asking for years, in rooms where people are paid to know the answer, and I have never once received a straight reply. The question is this:

Before a vaccine antigen is injected into a human being for the first time, is anyone required by law to check whether it looks like a human protein?

The answer is no. Not “it depends.” Not “in most cases.” No. There is no federal statute and no federal regulation that requires it. I went through the statutory chain and the implementing regulations line by line, and the gap is not a gap in enforcement or a gap in guidance. It is a gap in the text.

That finding is worse than the one I went looking for, and it is also far more useful, because it is not arguable.

What the law actually says

Start with the statute. Section 351 of the Public Health Service Act, 42 U.S.C. § 262, conditions licensure of a biological product on a showing that “the biological product that is the subject of the application is safe, pure, and potent.”

Read the object of that sentence. It is the biological product. The finished article in the vial. Not its components. Not its antigen. Not any structural feature within the antigen.

There is a small irony buried in the same statute. The definition at § 262(i)(1) enumerates the regulated classes: “a virus, therapeutic serum, toxin, antitoxin, vaccine, blood, blood component or derivative, allergenic product, protein, or analogous product.” Congress named protein. It simply never required that a protein placed inside a vaccine be evaluated on protein-specific grounds.

Now the regulations. The operative definition of safety is at 21 CFR 600.3(p):

The word safety means the relative freedom from harmful effect to persons affected, directly or indirectly, by a product when prudently administered, taking into consideration the character of the product in relation to the condition of the recipient at the time.

That is the whole thing. It is relative, it is outcome-framed, and it contains no composition requirement of any kind. Its companion definitions do not help. Purity at § 600.3(r) is “relative freedom from extraneous matter in the finished product” — a requirement about what should not be in there. Potency at § 600.3(s) is about whether the product works. Nowhere in the triad does anything ask whether the intended ingredient is itself hazardous by virtue of its structure.

Lot release does not fill the gap either. 21 CFR 610.1 forbids releasing a lot before completion of tests for conformity with applicable standards — conformity testing, not characterization. The only identity requirement, 21 CFR 610.14, tests the contents of a final container “for identity after all labeling operations shall have been completed.” That confirms the vial holds what the label says. It does not ask whether what the label says is safe to inject.

Preclinical requirements are at 21 CFR 312.23(a)(8): pharmacology and toxicology information on the basis of which the sponsor has concluded it is reasonably safe to proceed. Sponsor-determined scope. Nothing specifying analysis of the immunogen.

And then there is the detail that will tell you everything about how this field thinks.

The section that is now blank

Pull up 21 CFR 610.11 today. It is captioned nothing. It reads [Reserved].

It used to be captioned “General safety test.” It was revoked, along with § 610.11a and § 680.3(b), by a final rule published at 80 FR 37971 on July 2, 2015, Docket FDA-2014-N-1110, effective August 3, 2015. The rule’s own title states the rationale: Revocation of General Safety Test Regulations That Are Duplicative of Requirements in Biologics License Applications. The conforming cross-references in §§ 601.2(c)(1) and 601.22 were struck at the same time.

I want to be fair about this, because the reform case is weakened every time someone overstates it. The revocation did not abolish anyone’s obligation. FDA’s position was that the requirement lives in each product’s license rather than in the published rule, and the animal-welfare rationale for not running a non-specific abnormal-toxicity test in guinea pigs and mice forever was genuine. I do not think the agency acted in bad faith.

But look at what happened structurally. A safety requirement that was published, uniform, and citable became a set of product-specific license commitments that are not public and not uniform. You can no longer open the Code of Federal Regulations and determine what general safety testing any particular vaccine underwent. You have to take it on faith that the commitment exists in a document you cannot read.

And here is the part that matters most for the argument I am making: even when § 610.11 was fully in force, it was a non-specific abnormal-toxicity test in animals. It was never a characterization of the antigen. The provision whose loss people point to would not have supplied what is missing. The thing I am asking for has never existed in American law at any point.

The asymmetry that should end the argument

Here is where this stops being a complaint about a gap and becomes an indictment of inconsistency.

If you develop a therapeutic monoclonal antibody, FDA expects you to assess whether that antibody binds human tissues other than its intended target before you put it into a person. The expectation is reflected in the ICH harmonised guidance S6(R1), Preclinical Safety Evaluation of Biotechnology-Derived Pharmaceuticals. In practice it means an immunohistochemical screen against a panel of normal human tissues. The scientific reasoning is obvious and nobody disputes it: an immunoglobulin introduced into a human body may engage things it was not designed to engage, and where it engages them is material to safety.

Now consider what a vaccine does. A vaccine does not deliver an antibody. It instructs the recipient to manufacture antibody, and to maintain immunological memory capable of manufacturing more of it for years or for life.

In both cases, immunoglobulin ends up in the patient. In the monoclonal case, it is a finite dose with a known half-life, and you screened it against human tissue first. In the vaccine case, it is self-renewing and durable, and nobody screened anything.

I have asked, repeatedly, for the scientific principle that justifies this. Not the regulatory history — the principle. Why does a passively transferred antibody warrant a human tissue cross-reactivity panel while an actively induced one, which is harder to recall and lasts far longer, warrants none?

No one has given me one. I do not think one exists. What exists is a historical accident: vaccines were regulated as biologics in 1902 under a framework built around contamination and potency, monoclonals were regulated decades later under a framework built around molecular specificity, and the two expectations never reconciled.

This is not a hypothesis about autoimmunity. It is settled immunology.

I want to be precise here, because this is exactly the point at which a reform argument usually gets dismissed, and usually deserves to be.

I am not claiming that vaccines generally cause autoimmune disease. I am not claiming that homology between a vaccine antigen and a human protein means a product is dangerous. I will go further and say that anyone who claims that is wrong, and I will explain below exactly why they are wrong.

What I am saying is narrower and is not contested by anyone in immunology:

The immune system’s capacity to mistake self for non-self is a documented cause of human disease, and the structural basis of that mistake is sequence and conformation.

Two examples establish it beyond argument.

Acute rheumatic fever. Immune responses to Streptococcus pyogenes M protein cross-react with cardiac myosin and related determinants in heart valve tissue. Weeks after a strep throat, in a susceptible host, an immune response aimed at a bacterium damages a heart. This has been understood for decades. It is in every immunology textbook.

Guillain-Barré syndrome after Campylobacter jejuni. Antibodies raised against C. jejuni lipo-oligosaccharide structures cross-react with gangliosides in peripheral nerve, producing acute motor axonal neuropathy. Same mechanism, different tissue, different molecular class.

Nobody disputes either one. They are not fringe findings; they are the foundation of what we teach about molecular mimicry.

Now add the obvious observation. Sequence and conformation are knowable before anything is injected into anybody. They are knowable before the first vaccine is made. A microbial antigen’s amino acid sequence is a text file. The human proteome is a text file. Comparing them is an afternoon’s work on a laptop, using public reference databases and public tools. Predicting which regions are likely to be epitopes — recognized by B cells, or presented on MHC molecules across common human HLA alleles — is now routine computational immunology with publicly maintained reference tools. Looking up which human tissues express a matched protein is a database query.

This is the core of my pathogenic priming analysis, published on the SARS-CoV-2 viral proteins in April 2020. I had hoped it would help vaccine makers avoid inclusion of unsafe epitopes.

It did not, and, as a result, chronic illness following COVID-19 mRNA jabs abounds.

We have the mechanism. We have the method. We have had both for years. And we have no requirement.

Why a naive version of this demand would fail, and should

Here is the part that most people arguing for this leave out, and leaving it out is why they lose.

Short peptide matches between any two large proteomes are everywhere. Twenty amino acids, a five-residue window, and two proteomes of tens of thousands of proteins each — matches are a combinatorial certainty. They are expected. They are, in the overwhelming majority of cases, immunologically silent.

Homology is neither necessary nor sufficient for cross-reactive pathology. Not sufficient, because the vast majority of linear matches never produce a detectable autoreactive response, let alone clinical disease — the matched region may be buried in the folded protein, never presented, never seen. Not necessary, because cross-reactivity can arise from conformational similarity between sequences that share no linear identity at all.

So if you demand a rule that treats any homology as a safety signal, you will produce an enormous pile of findings that mean nothing, you will impose cost without yield, and you will make safety decisions worse rather than better. You will also hand every opponent of the proposal a free win, because they will be right.

The rule has to be built the way a competent scientist would build it:

Tier 1 is characterization and disclosure, not pass or fail. Run the comparison. Annotate every match. Say which human proteins are implicated and which tissues express them. Run the identical analysis against control sequences so the match count can be read against the expected background instead of floating free. Report it all, publicly. Nothing is approved or rejected on the basis of Tier 1. It is a map.

Tier 2 fires only on a conjunction. Confirmatory wet-laboratory testing is triggered when a match satisfies both conditions: it falls within a demonstrated or predicted epitope, and the matched human protein is expressed in a tissue where autoreactivity would matter — heart, central nervous system, peripheral nerve, pancreatic islet, thyroid, kidney, platelet, vascular endothelium, hematopoietic lineage. That conjunction does the work. It takes the uninformative pile and reduces it to the handful of findings that could plausibly mean something.

Tier 2 uses methods that already exist. Peptide binding assays against pre- and post-immunization sera. Immunohistochemistry against a human tissue panel — the same format already run for monoclonals. Functional assessment where binding is demonstrated. None of this requires invention. It requires application.

And the rule says, in its own text, that homology is not evidence of harm. It has to say so explicitly, because otherwise the first published report showing a match gets misread by people on my side of this argument as proof of something, and the entire exercise is discredited within a week.

And we need a clarification on short peptides: the epitopes themselves are the subject of homology consideration, and, as a result, we can expect insignificant E-values from a protein blast even when the result is, in fact, biologically significant. So the initial list will include many false positives.

The follow-up lab confirmation is necessary.

That last point is not a concession I am making to be diplomatic. It is the condition on which the proposal is scientifically honest. A screen that cannot come back negative is not a screen.

Where the legal authority already is

The most common objection to any proposal like this is that the agency lacks authority and Congress would have to act.

Congress already acted. In 1986.

42 U.S.C. § 300aa-27, “Mandate for safer childhood vaccines,” provides at subsection (a) that the Secretary shall:

(1) promote the development of childhood vaccines that result in fewer and less serious adverse reactions than those vaccines on the market on December 22, 1987, and promote the refinement of such vaccines, and (2) make or assure improvements in, and otherwise use the authorities of the Secretary with respect to, the licensing, manufacturing, processing, testing, labeling, warning, use instructions, distribution, storage, administration, field surveillance, adverse reaction reporting, and recall of reactogenic lots or batches, of vaccines, and research on vaccines, in order to reduce the risks of adverse reactions to vaccines.

Subsection (b) establishes a task force on safer childhood vaccines — the Director of NIH as chairman, the Commissioner of FDA, the Director of CDC — to prepare recommendations on implementing subsection (a). Subsection (c) requires reports to the House Committee on Energy and Commerce and the Senate committee of jurisdiction describing what has been done.

Read what that provision contains. It is affirmative — “shall,” not “may.” It is a standing duty on the Secretary, not a condition a sponsor discharges and walks away from. It names testing. It names research on vaccines. It ties both to the express purpose of reducing the risks of adverse reactions. And it names refinement of vaccines already on the market, which is precisely what removing a problematic epitope from a licensed antigen would be.

A cheap, pre-clinical, computational screen that identifies — before first-in-human administration — antigen features associated with a documented mechanism of adverse immune reaction is not adjacent to that mandate. It is the center of it.

So the honest framing of this problem is not “the law requires antigen screening and FDA is ignoring it.” That claim is false and would collapse under the first competent rebuttal. The honest framing is this:

Congress imposed a standing duty to improve vaccine testing and research in order to reduce adverse reactions. The screening I am describing is cheap, validated, and squarely within that duty. Forty years on, no regulation implements it.

That statement is accurate. It names an addressee who already holds the obligation. And it cannot be answered by pointing to a provision, because there is no provision to point to.

Why this should extend to biologics generally

I have written this around vaccines because that is where the mandate language is strongest and where the exposed population is largest and youngest. But the logic does not stop there, and neither should the rule.

Any biologic that places a protein, or protein-coding instructions, into a human being raises the same question. Gene therapy vectors express transgene products. Nucleic acid platforms express whatever the construct encodes — and, worth asking, whatever alternative reading frames in that construct encode, which is a characterization almost nobody performs and nobody is required to perform. Engineered enzymes and fusion proteins carry linkers, tags, signal sequences, and stabilizing substitutions that exist in no natural proteome, which means their homology profile has never been tested by evolution against any host.

That last point deserves emphasis. A wild-type microbial protein has at least been through a long history of contact with human immune systems, and the catastrophic mimics have mostly been selected against — we notice when they have not been, which is how we learned about rheumatic fever. An engineered sequence has no such history. A stabilizing proline substitution, a linker chosen for flexibility, a tag left in from purification — each creates junctions and surfaces that have never existed before. The argument for pre-administration screening is strongest exactly where the sequence is most novel, and novelty is the direction the entire biologics industry has been moving for thirty years.

What this costs

Almost nothing, and that is the part I would like opponents to have to argue about in public.

The Tier 1 analysis is computational. It uses public reference proteomes, public epitope prediction tools, public tissue expression databases. The cost per antigen is analyst time. There is no animal, no reagent, no facility.

Tier 2 costs more, but Tier 2 is bounded by the conjunction trigger, and the assay formats are the ones already validated and in routine commercial use for therapeutic antibody development. Nothing has to be invented or qualified from scratch.

This is the unusual safety proposal where the burden analysis favors the people asking for it. Which suggests a test anyone can apply: run the Tier 1 screen retrospectively, on every currently licensed vaccine antigen, and publish the results. Nothing prevents NIH or FDA from doing it this quarter. If the screens come back clean, the field gets a durable, public, auditable answer to a question that has been asked in bad faith for twenty years, and the people asking it lose their best talking point. If they do not come back clean, we needed to know.

I notice that nobody who dismisses this concern has proposed running the test. That is worth sitting with. If you are confident the answer is clean, the cheapest way to win the argument permanently is to produce the data.

What I am asking for

The U.S. CFR needs a rule, through ordinary notice-and-comment rulemaking, requiring that the sponsor of any vaccine or protein-expressing biologic:

1. Submit, with the original IND and before first administration to a human being, a full computational characterization of the antigen’s homology to the human proteome, annotated for predicted and demonstrated epitopes and for the tissue expression of every matched human protein, run alongside controls, with full version and parameter disclosure so the analysis can be exactly reproduced.

2. Characterize the actual translated products for nucleic acid and vectored platforms, including alternative reading frames and truncated species.

3. Conduct confirmatory peptide, tissue cross-reactivity, and functional testing where, and only where, a match falls in an epitope and implicates a sensitive tissue.

4. Have the analysis performed or independently replicated by a party with no financial interest in the product.

5. Make the whole report public. The antigen sequence of a licensed vaccine is derivable from public information already; there is no trade secret here worth the price of secrecy, and anyone claiming otherwise should have to say which part and why, on the record.

6. Do all of the above retrospectively for currently licensed antigens, prioritizing the childhood schedule — which is what the statute itself prioritizes.

And a rule that states plainly, in its own operative text, that identification of homology is not evidence that a product is unsafe and is not alone a basis for withholding or revoking a license.

That last clause is not a hedge. It is the clause that makes the rest of it true.

The sentence I keep coming back to

We require a test confirming that the vial contains what the label says. We require a test for extraneous matter — for what should not be in there. We require evidence that the product produces the intended immune response.

We require nothing about the one component we put in on purpose, for the express purpose of provoking an immune response, chosen specifically because the immune system will react to it.

That is not a conspiracy. It is something more ordinary and more correctable: a framework built in 1902 around contamination and potency, extended across a century of molecular biology, that was never updated to ask the molecular question. The tools to ask it have existed for two decades. The statutory duty to ask it has existed since 1986. The cost of asking it is close to zero.

All that is missing is the requirement. And requirements are written by people.

Yesterday, sent a detailed rule-making plan to Secretary Kennedy that requires the testing and screening of biological molecules and compounds in vaccines to be tested for their capacity to induce autoimmunity via pathogenic priming.

James Lyons-Weiler, PhD is founder and president of IPAK-EDU LLC and Editor-in-Chief of Science, Public Health Policy, and the Law.

The research facilitation and study design tools referenced in this analysis are part of the IPAK-EDU research suite. Courses and webinars on study design, bias detection, and regulatory analysis are available at

Home

— and a paid subscription to Popular Rationalism supports work like this.

Sources

Lyons-Weiler, 2020. https://www.sciencedirect.com/science/article/pii/S2589909020300186

See Cited By:

https://www.sciencedirect.com/science/article/pii/S2589909020300186?via%3Dihub#cited-by

42 U.S.C. § 262 — Regulation of biological products — https://www.law.cornell.edu/uscode/text/42/262

42 U.S.C. § 300aa-27 — Mandate for safer childhood vaccines — https://www.govinfo.gov/content/pkg/USCODE-2016-title42/html/USCODE-2016-title42-chap6A-subchapXIX-part2-subpartc-sec300aa-27.htm

21 CFR 600.3 — Definitions — https://www.law.cornell.edu/cfr/text/21/600.3

21 CFR 610.1 — Tests prior to release required for each lot — https://www.law.cornell.edu/cfr/text/21/610.1

21 CFR 610.14 — Identity — https://www.law.cornell.edu/cfr/text/21/610.14

21 CFR 312.23 — IND content and format — https://www.law.cornell.edu/cfr/text/21/312.23

80 FR 37971 — Revocation of General Safety Test Regulations, July 2, 2015 — https://www.federalregister.gov/documents/2015/07/02/2015-16366/revocation-of-general-safety-test-regulations-that-are-duplicative-of-requirements-in-biologics

FDA guidance search, for ICH S6(R1) — https://www.fda.gov/regulatory-information/search-fda-guidance-documents

 

IPAK-EDU is grateful to Popular Rationalism as this piece was originally published there and is included in this news feed with mutual agreement. Read More