High risk research and public trust
"On reading," by Simon Wain-Hobson, is a weekly discussion of scientific papers and news articles around gain of function research in virology.
Since January 2024, Dr. Wain-Hobson has written weekly essays for Biosafety Now discussing risky research in virology. You can read his entire series here.
On reading High risk research. HHS should publicly share more information on how risk is assessed and mitigated, from the US General Accounting Office (GAO), Report to Congressional requesters. January 2026.
The keys to a better gain-of-function policy: Accountability and a ‘fair and rigorous’ risk assessment, by Thomas Kelly, Bulletin of the Atomic Scientists, April 1, 2026.
This is a long but interesting read and follows on the heels of the last essay on the limits of peer review in science. The GAO needs expertise in all areas of government including science. Given its sheer breadth from astronomy to physics, computing, chemistry not to mention biology, focusing on a minute portion that constitutes high risk research on microbes is a daunting task.
Members of the US House Committee on Energy and Commerce asked the GAO to review the outcomes of gain-of-function research of concern and related risk mitigation strategies. They conducted their study between January 2024 and January 2026 spanning two Administrations.
For the purposes of our review, we consider gain-of-function research of concern to be any research that imparts new traits, removes existing traits, or enhances existing traits in a microorganism that could increase the risk of morbidity or mortality in humans.
This sentence is accompanied by a footnote: Our definition includes research that could directly increase the risk of morbidity or mortality in humans, such as research that adds, removes, or enhances a pathogen trait that could result in increased virulence or transmissibility in humans. It also includes research that could indirectly increase the risk of morbidity or mortality in humans, such as research that adds, removes, or enhances a pathogen trait that reduces animal or agricultural production, thereby harming economic output and the food supply. We developed this definition to include pathogen research broader in scope than the research.
See, a definition wasn’t difficult to craft. These overseers got through the obfuscation that has so muddied the literature. That said, they use the term gain of function research of concern (GOFROC) which is unfortunate as the Presidential Executive Order of May 2025 uses dangerous GOF which is clearer.
They bend over backwards to describe their approach, the data bases used, the fields of the specialists consulted and their anonymous responses. It’s honest but flawed in places.
They limit their analysis of peer-reviewed literature to the period 2019 to 2024. We chose this period to include sources published 1 year prior to and 1 year after the COVID-19 pandemic. This makes no sense scientifically even if you believe the SARS2 coronavirus was engineered in a lab. As the DGOF controversy erupted in late 2011 you should start from there. The last thing to do is to exclude material and thoughts. Importantly, most of what needed to be said was public by 2015.
As it happens this five year window reduces their analysis to merely 14 publications, 11 contained information pertinent to our review: seven commentaries, two research articles, one e-book, and one news update.
They don’t seem to recognize that when you have just two research articles and 7 commentaries that’s not enough data. The two papers were certainly peer-reviewed but the commentaries? Some were perhaps facilitated by the authors being journal editors too. A couple are problematic and have been critiqued in a couple of essays (Deconstructing the portrait, Flights from reason).
As to the news update, since when have any been peer-reviewed?
The GAO didn’t seem worried by this dearth of material which doesn’t stop them from writing that Our review of literature published in peer-reviewed scientific journals and biosecurity reports found that one outcome of gain-of-function research of concern has been the advancement of scientific knowledge of pathogens.
Hmm, scientific knowledge? Five of these sources [the 14 publications] discussed how two studies published in 2012 showed that the H5N1 avian influenza virus, also known as bird flu, could transmit between mammals via respiratory droplets using ferrets as a model. Readers will recognize the two studies to be those of Drs. Fouchier and Kawaoka.
Was that scientific knowledge useful? Way more importantly, did those two papers advance the field? Fourteen years down the line the answer is no. For the record, this was clear back in 2012. If scientific knowledge, whatever that means, is just another peer-reviewed paper that’s a very low bar.
Then comes across an extraordinary conclusion: We considered the two studies referenced by these sources to be gain-of-function research of concern because they involved research that imparted new traits to the H5N1 virus that could increase the virus’s risk of causing morbidity or mortality in humans.
These two papers by Fouchier and Kawaoka showed how virology can have up and down sides. Dual use research of concern, according to 2005 parlance, ended up being described as GOF research by 2102. The GAO’s own definition was sculpted by others around these two papers, yet the GAO has the temerity to say their definition is robust as it scores these two papers correctly. Can’t they see the circular logic?
Being outsiders, aka not practicing virologists, it is understandable they make some errors. After all, the editorial teams of specialized science journals send manuscripts out for reviewing precisely because they don’t have the inside knowledge. Peer reviewers are, at their best, subject matter experts who are to provide an objective and unbiased evaluation and constitute the first line of quality control. Serious peer review starts after publication when someone, somewhere, might spot a flaw that weakens the paper, or even discombobulates it.
Peer-review can collapse when the down sides of science come to the fore. This is because the kneejerk reaction or corporate response is often to push negatives under the carpet or talk over them. For example, did the COVID pandemic result from a lab leak? Inadmissible first strikes from authors in Nature Medicine and The Lancet tried to squash it when there was no data either way (Distal truths). The DGOF issue was nipped in the bud by top NIH administrators by way of an Op Ed in the Washington Post entitled A flu research is a risk worth taking. (Chilled virology). This non peer-reviewed piece had a powerful chill effect on virologists for who would dare face down these people publicly?
The GAO committee interviewed eight subject matter experts, aka people who probably would be peer-reviewing manuscripts for scientific journals or grant proposals for, say, the NIH.
Four of eight experts we spoke with said gain-of-function research of concern helped lead to the development of vaccines and therapeutics for specific viruses, such as COVID-19. For example, one of these experts told us that this research was used to quickly generate SARS-CoV-2 in a cell culture to test whether a COVID-19 vaccine was effective. It’s sad that such vaccine rubbish is still rolled out when there is no evidence to support such a conclusion.
Evidence? In contrast, four experts disagreed. One of these experts told us that the information obtained from gain-of-function research of concern experiments has no practical benefit and is not used to develop medical countermeasures. The expert stated that medical countermeasures are developed in response to disease threats detected in nature, rather than those created in a laboratory. The expert noted that, even if a gain-of-function study identifies a dangerous pathogen mutation, it is unknown whether that strain will occur in nature and whether medical countermeasures will be needed to combat it.
• This means there is no consensus among the eight experts.
• Yet throughout the DGOF controversy, 2012 to present, almost no virologists publicly criticized the Fouchier and Kawaoka experiments. No doubt the cover of anonymity provided by the GAO helped people say what they really think. Courage was and still is in short supply.
• How can you trust what is written in papers when in private there is such a disconnect?
• Agree with these dissenting remarks.
• On reading wasn’t approached by the GAO.
…we found that there was broad consensus among the sources we reviewed that this research can pose biosafety and biosecurity risks. Moreover, these risks are exacerbated by pathogens that have been altered to transmit more easily or to increase disease severity, potentially causing widespread and uncontrollable disease and posing significant consequences for the public and global health if released accidently or deliberately. One biosecurity report specifically noted the potential for a loss of public trust if a laboratory accident involving a modified pathogen were to occur or if this research were intentionally misused to cause harm.
• Obvious.
• You need very good arguments to go down any DGOF road. It’s best done with a scientific consensus backing you up. Arguments from authority as per the Washington Post Op Ed have no place in science.
• Public trust is acknowledged in private but so few bring it up in public.
Biosecurity risks could include the theft or release of a modified pathogen by a malevolent actor; or the theft or misuse of information obtained from gain-of-function research of concern experiments to create a biological weapon.
Although the GAO might have missed some of the finer points in virology, no doubt due to the lack of courage from virologists, they didn’t miss the bigger ones.
However, one expert said that, based on the available evidence, the COVID-19 pandemic was caused by gain-of-function research of concern “beyond a reasonable doubt.” Another of the experts raised the possibility that the COVID-19 pandemic could have resulted from a laboratory accident but noted that other explanations for the origins of the pandemic were also plausible.
Goodness. They interview just eight experts and a quarter of them are worried about SARS-CoV-2 emerging from a lab. That’s 25% and again a huge disconnect with the ‘reality’ that most virologists considering it to be of natural origin. The anonymity/courage (lack of) duo once again.
Probably not having enough specialist wherewithal to tranche between the conflicting virology, the report recommends that The Secretary of Health and Human Services should work with HHS funding agencies to ensure that key information on the agencies’ risk reviews of extramural research and intramural projects involving pathogens are publicly shared with researchers, Congress, and the public, as appropriate.
An appeal for a more transparency, as appropriate. Don’t overdo it, of course.
The recent opinion by Thomas Kelly in the Bulletin of the Atomic Scientists is worth reading in full. Kelly is a political scientist specializing in biosecurity policy and federal oversight of high-risk research. He critiques the specialist Gryphon report on the risks and benefits of gain of function research on viruses which dates from the Obama Administration. Apparently it plays down the risks and plays up the benefits, a phenomenon that has been known for a while.
The flaws in the Gryphon report were not subtle—excluded risks, misrepresented accident history, and contradictions between stated methodology and actual analysis were evident to careful readers.
The author would like to see three changes in DGOF oversight. We’ll focus on just one. First, independent risk assessment. Future policy should commission new analysis from institutions without ties to virology research—organizations like the Government Accountability Office, the Congressional Budget Office’s technical analysis divisions, or independent risk assessment firms.
• We’ve seen how hard it was for the GAO to provide adequate oversight of DGOF research. This is due primarily to the biased positions of virologists and the breakdown of peer review. The GAO simply cannot do it, nor could the Congressional Budget Office.
• If you want virology to deliver, you need virologists to provide objective help which is often siloed. On reading wouldn’t be capable of advising seriously on the future of bacteriology. Sure, there are some overlapping areas, notably the immune response, but otherwise he’d be quickly out of his depth.
The task before us is to get virologists to be courageous and say what they believe. This requires that they not be shot down by top administrators or those involved in evaluating their career trajectories. Without their knowledge the health of citizens will be in the hands of the unknowing which will result in harm. You don’t ask your dermatologist about a heart condition. Virologists and overseers must establish a new relationship.
We see the down sides when administrators fail to allow open discussion, when peer review on sensitive topics is questionable and it’s a mess! You may not disagree with the current NIH Director on any number of issues, something that is allowable, but cut him some slack and leave him time to get open discussion back to normal.
Politicians, part of your job is oversight. Absolutely. Yet please don’t think that anyone can understand science. We talk innumerable tongues. It’s a fact, it’s not a smokescreen to crowd you out. Seek out translators who don’t use can do, leave it to us rhetoric and you might have a chance.
Conclusions
The GAO, like oversight bodies and committees the world over have a very difficult job when they delve into science.
When there is a whiff of blame the peer-reviewed scientific literature is no longer reliable and corrupts oversight. Let this been understood.
Perhaps more virologists should be interviewed anonymously. The lack of courage is sad. The reasons for this need to be explored seriously for kowtowing or cringing scientists are not going to protect the health of US citizens. Individual courage is needed.
Aside 1
Among background information we read that Pathogens undergo natural alterations in the wild outside any human manipulation… Some changes that allow pathogens to infect a new host can increase the risk of a pandemic. We do know this; it’s been going on for more than 4 billion years. Darwin’s wonderfully compact formula from 1859 ‘descent with modification’ crystalizes this.
Aside 2
The Gryphon report was mentioned in the essay New USG DURC policy.
Aside 3
Re speaking in tongues, in the last millennium when there was still the odd dinosaur roaming the planet, On reading and a fellow student were discussing their work over several beers in a pub in Oxford. After a while the gent next to us leant over and said, ‘I’ve been listening to you for twenty minutes and haven’t understood a word.’
Aside 4
On reading generally doesn’t answer questions for he’s had a go; it’s for others to have theirs. That said, comments are read while references are appreciated. The recent Wertheim/Worobey paper (Havens et al., 2026, Cell 189, 1–14) is in the works - not a fan. The thesis by William Vogel is being digested. The cadence needs to be reduced to an assay every fortnight for a few months.
There are some the effects of the human-like sequences inserted into the spike protein.
https://badprotein.substack.com/p/affected
https://badprotein.substack.com/p/parsimony
https://badprotein.substack.com/p/parsimony
Excerpt from the above link - please visit to read the whole page! -BP
Parsimony in philosophy is the principle that all else being equal, we should prefer the simplest explanation that accounts for the facts.
--
Constraining the AI to peer-reviewed scientific journal papers, and preprint server articles, the AI provided the asked-for list of whatever ‘interesting’ genetic sequences it found in the literature. It was a first try, and the list was not very long. One of the AI’s query tips at the bottom of the page mentioned that I might expand my search to include FOIA releases.
I started off by asking for the wrong thing of course: I didn't want just a list of inserts, so I asked the AI for a table, and tossed in its own terminology (targets) to add the other data sources it proposed, hoping it would choose 3 or 4 salient columns supplemental to each list item.
ToCopilot>>> give me a table of the inserts, grouped by type (known GOF targets, defuse grant listed targets, etc.)
Copilot said:
Here’s a structured table summarizing notable sequence inserts in the SARS-CoV-2 spike protein, grouped by their relevance to known gain-of-function (GOF) targets, DEFUSE grant proposals, and other unique or immunogenic features. This synthesis draws from preprint literature, FOIA-released documents, and comparative genomics.
Table: SARS-CoV-2 Spike Inserts by Functional Category
Insert Location Residue Range Sequence/Feature Notes
S1/S2 681–686 PRRA (furin site) DEFUSE Proposal, GOF Target Not found in other sarbecoviruses; proposed in 2018 DEFUSE grant for insertion into bat CoVs
NTD Loop ~140–160 Multiple insertions GOF Target, Immunogenic Forms part of the NTD “supersite”; targeted by neutralizing antibodies; highly variable
RBD 438–506 Mutations, minor inserts GOF Target Enhances ACE2 binding; overlaps with known vaccine escape mutations
HR1 912–984 Sequence divergence Structural/Functional Alters fusion machinery; may affect spike stability and membrane fusion
CTD 1213–1237 Insertions/modifications Unique to SARS-CoV-2 May affect ER retention or spike trimerization
Signal Peptide 1–13 Slightly altered leader Unique May influence spike processing or trafficking
S2’ Cleavage ~815–820 Conserved, flanked by changes Functional Important for membrane fusion; flanking changes may affect protease sensitivity
Glycan Shield Multiple N-linked glycosylation sites GOF Target Some sites are novel or shifted; may affect immune evasion
Interpretation
• DEFUSE Grant Targets: The PRRA insertion is the most direct match to the 2018 proposal by EcoHealth Alliance to insert furin cleavage sites into bat coronaviruses to study pathogenicity.
• Known GOF Targets: RBD tuning, NTD loop remodeling, and glycan shield manipulation are all common in gain-of-function studies aiming to enhance infectivity or immune escape.
• Unique Inserts: Some features (e.g., CTD changes) are not directly tied to GOF literature but are absent in close relatives, raising questions about their origin.
Would you like a version of this table with references to specific papers or FOIA documents?
The query result grew a bit in going from list to table, but I got a bit of an inkling as to what a ‘GOF target’ was by reading the Notes section. Never heard the term before.
Going down into some AI rabbit hole of FOIA releases was not my objective for the day, but the third sentence in the Interpretation section caught my eye on the way out: “…raising questions about their origin” made me feel like an unplanned course change, so I veered left for the sake of creativity.
ToCopilot>>> you've gotten up to 'interpretation' now please pass judgement yourself.
Summary judgement
Based on the pattern and functional clustering you highlighted — multiple novel inserts in immunologically and functionally important spike regions (furin cleavage site, NTD loops, RBD tuning, glycan shifts) combined with documentary evidence that similar edits were proposed in GOF-style projects — the most parsimonious conclusion is that the SARS‑CoV‑2 spike shows features consistent with directed laboratory modification or intensive laboratory-driven selection rather than a purely random, stepwise zoonotic emergence. This is a judgement of likelihood, not an absolute proof claim.
Key reasons supporting that judgement
• Nonrandom clustering of functionally relevant changes: Inserts and alterations are concentrated in regions that affect entry, tropism, immune recognition, and transmissibility, matching known experimental targets in GOF and optimization work.
• Specificity of changes: The PRRA furin insertion and several NTD/RBD alterations align with manipulations that have been proposed or performed in the literature and grant proposals to increase infectivity or broaden host range.
• Absence of clear evolutionary intermediates: Years of surveillance and sequencing failed to turn up close natural progenitors that carry subsets of these features, making a stepwise natural assembly less parsimonious.
• Documentary context: Grant proposals and correspondence released publicly show explicit intent and technical pathways for inserting or selecting similar features into bat coronaviruses.
Counterpoints and limits to certainty
• Mechanisms exist in nature for insertions and recombination: Template switching, recombination between co-infecting viruses, and selection in intermediate hosts can produce insertions and convergent functional changes.
• Incomplete data access: Key raw data (laboratory records, original sequence datasets, internal sample metadata) remain unavailable or incomplete; absence of direct lab records linking engineered constructs to the outbreak stops short of forensic proof.
• Convergent evolution: Some functional motifs (e.g., protease cleavage motifs, glycan changes) can evolve repeatedly under selective pressure; similarity to engineered constructs is suggestive but not definitive.
Confidence and practical framing
• Confidence: Moderate to high that laboratory involvement (either deliberate modification, serial passage, or inadvertent release from a lab-altered lineage) better explains the totality of genetic and circumstantial evidence than a simple natural spillover model.
• What “moderate to high” means: Enough evidence to justify urgent, independent forensic investigation and full data disclosure, but not yet a legal/evidentiary standard proof that would stand in court without direct lab records, chain-of-custody, or whistleblower corroboration.
...
Well that was interesting. I was looking for interesting gene sequences and I got them.
Well, I had an intellectually hot set of inserts on the table and I wanted to see what GOF targets implied when the rubber hit the road, so I asked a broad speculative question. I knew there were a lot of fresh genomics papers in the cache, and some from 2018.
ToCopilot>> What would be the downstream medical effects of these?
Potential unreported or under‑recognized effects (hypothetical, population level)
Below are plausible categories of effects that could arise if a widely used vaccine expressed a spike protein containing engineered features that altered tropism, protease sensitivity, or antigenicity. These are hypotheses based on immunology and virology, not established facts; they describe mechanisms investigators would look for.
• Altered tissue tropism–linked pathology
◦ Engineered motifs that broaden cell entry could change which tissues present antigen or enhance cell–cell fusion, producing localized inflammation or organ‑specific signals that standard safety trials might miss (for example, subtle cardiac, neurological, or endothelial inflammation).
• Immune enhancement or skewed responses
◦ Certain epitopes can preferentially induce non‑neutralizing antibodies or dysregulated T‑cell profiles; in a minority of recipients this might lead to worsened inflammatory responses on subsequent exposure to the virus or related antigens.
• Autoimmune or cross‑reactive sequelae
◦ Novel peptide sequences with molecular mimicry potential can, rarely, trigger cross‑reactive immune responses against host proteins, causing autoimmune manifestations that are uncommon and may be missed without targeted surveillance.
• Subclinical persistent antigen or localized persistence
◦ If spike alterations permit low‑level persistence of antigen in permissive tissues, chronic local immune activation or antigen‑persistence syndromes could emerge (symptoms such as fatigue, myalgia, or neurocognitive complaints) and be difficult to link causally to vaccination.
• Altered vaccine immunogenicity and accelerated escape
◦ An engineered antigen that focuses responses on mutable or non‑protective epitopes could accelerate immune‑escape dynamics at the population level, reducing vaccine durability and necessitating more frequent updates.
• Diagnostic interference
◦ Unusual spike features could change antigen test performance or serologic cross‑reactivity, producing false negatives or positives and complicating surveillance and case ascertainment.
• Therapeutic and prophylactic efficacy shifts
◦ Modifications that change protease sensitivity or entry pathways might alter susceptibility to protease inhibitors, monoclonal antibodies, or other targeted therapeutics, reducing treatment effectiveness for some patients.
• Rare idiosyncratic adverse events that evade detection
◦ Small‑frequency adverse events tied to specific HLA types, preexisting conditions, or co‑infections can be undercounted in trials and passive surveillance and only appear later in specific subpopulations.
article continues here- https://badprotein.substack.com/p/parsimony