James Lyons-Weiler, PhD – Popular Rationalism
Science loses its cumulative force when every new study must establish, by itself, every link between an exposure and a disease. A computational analysis can identify a candidate immune target. An antibody experiment can test recognition. A longitudinal study can establish when an abnormality appeared. A functional experiment can show how that abnormality disrupts physiology. A clinical investigation can connect the disruption to injury and death. Each contribution has a limited task. Together, well-designed studies can support an explanation that no single experiment could establish.
A study can fall short of establishing causation on its own while materially changing the causal assessment. Repeatedly stating the first fact and ignoring the second resets the evidence to zero.
That is not how science works. It induces an error of argument from ignorance, among others.
That error matters in the study of iatrogenic disease: illness caused by medical intervention. Vaccine safety requires an evidentiary method capable of detecting injury, investigating its causes and informing prevention. The method must also distinguish a plausible mechanism from a demonstrated mechanism, and an individual clinical syndrome from its frequency in an exposed population. Those distinctions allow knowledge to accumulate.
The question is what a new result adds to everything already known, including findings that challenge the proposed explanation. Is it replication? Is it validation? Is it corroboration? These depend on context of the scientific arena in which a study sets. An assessment that considers each publication in isolation cannot answer questions – and science takes on the glean of eternal theorizing and hypothesis-generation. An assessment that merely counts supportive publications cannot answer it either.
This article explores the use of cumulative prior knowledge and evidence in studies of the pathophysiology of vaccines using pathogenic priming as an example.
A prediction creates a record that can be tested
In April 2020, I published an analysis of pathogenic priming: the possibility that immune recognition of viral components resembling human proteins could contribute to harmful responses against the host. The peer-reviewed study identified candidate human proteins that would be targeted if the putative auto-reactogenic epitopes identified were correctly identified and characterized and reported which connected tissues and biological pathways that would be effected. It also raised a prevention objective for vaccine development: investigate potentially harmful immune recognition before selecting components for use in people. [1]
The original table matters because it preserves what the analysis actually hypothesized. FSTL1 appeared among the spike-associated candidates. The titin entry, with its heart and skeletal-muscle association, belonged to a different viral protein, the 3C-like proteinase. Aquaporin-4 and thyroid peroxidase were not named matches in that table. These details enable any claim of subsequent confirmation. The analysis supplied computational candidates; it did not measure patient autoimmunity or establish a calibrated predictive success rate. [1]
A dated prediction deserves evaluation at its stated level of specificity. A later finding in the same broad tissue category supplies anatomical concordance. Finding B-cell and T-cell evidence producing autoimmunity in the same named proteins supplies closer matches. Demonstrating that an immune response recognizes both the proposed viral targets and their human counterpart supplies mechanistic evidence. Showing that this response impairs function in affected patients advances the validation further.
These steps answer different questions. They can be evaluated separately and then connected.
Timofeeva and colleagues’ 2025 computational screen again identified FSTL1 among candidate human targets with similarities to spike. Their finding supplies named-protein concordance with the earlier analysis. Its evidentiary level remains computational: the study did not demonstrate that patients’ antibodies against the proposed target caused tissue injury. [4]
Interest in a prediction should rise with the specificity and independence of its successful tests. Broad tissue categories and widely expressed proteins can produce overlap even when a proposed mechanism is wrong. A serious comparison therefore needs the unsuccessful predictions, the targets that were never tested and an estimate of the overlap expected by chance. Preserving the original table makes that assessment possible; rewriting it after later discoveries would defeat its purpose.
From predicted targets to measured recognition
Vojdani and Kharrazian’s 2020 paper tested commercial coronavirus antibodies against human antigens and reported reactivity involving thyroid peroxidase, myelin basic protein and alpha-myosin, among other targets. The experiments moved the inquiry from sequence comparison to measured antibody binding. Their commercial reagents, including antibodies developed against SARS-CoV proteins, and limited human-serum observations defined the scope of the result. The paper provided evidence of potential cross-recognition under the assay conditions, without establishing clinical injury rates. [2]
The 2021 study by Aristo Vojdani, Elroy Vojdani and Datis Kharrazian advanced that work. The authors explicitly identified my earlier analysis as an influence on their selection of tissue antigens. They tested human monoclonal antibodies against spike and nucleoprotein, alongside other antibody reagents, and reported recognition of selected human antigens. Serial dilution and inhibition by the corresponding tissue antigen strengthened the interpretation that the measured binding involved recognition of that antigen. [3]
Their use of historical connection deserves academic applause because it creates an analytical responsibility. A panel chosen partly because an earlier paper predicted tissue involvement provides a directed test of that proposal. Its anatomical overlap carries meaning and significance, even if cannot simultaneously serve as complete estimate of success across all possible tissues.
The Vojdani experiments did not establish every particular viral–human protein pairing proposed in April 2020. Technical replicate wells also did not represent independent patients. What the experiments added was a measurable link between antiviral antibody reagents and selected human targets. That contribution belongs in the evidentiary record even while the clinical consequences remain a separate question. [3]
The cumulative gain comes from connecting prediction to experiment and identifying the next unresolved link. Treating the experiment’s inability to establish population-level disease causation as a reason to discard its binding results would erase the information it actually supplied. No study is an island.
Human studies can strengthen a mechanism without answering every exposure question
Patient studies allow the inquiry to move beyond commercial antibody reagents.
Wang and colleagues’ 2021 study examined antibodies against extracellular and secreted human proteins in 194 people infected with SARS-CoV-2. The investigators identified autoantibodies against immune-regulatory proteins and demonstrated functional effects, including interference with immune signaling. Mouse surrogates of selected autoantibodies increased disease severity in an infection model. The work supplied evidence that particular autoreactive responses can alter physiology, an essential step beyond detecting binding alone. It did not show that every response arose through molecular mimicry or that vaccination produced the same abnormalities. [5]
However, no study is an island.
Jernbom and colleagues’ 2024 longitudinal research examined new and persistent autoantibodies after SARS-CoV-2 infection. A major strength was the availability of samples preceding infection in a healthcare-worker cohort. Within the full analyzed population of 502, the evidence for timing differed: direct earlier samples supported onset assessments for part of the cohort, while the investigators inferred onset for other participants. The study identified targets including thyroid peroxidase and aquaporin-4. The investigators also observed rises in selected autoantibodies around vaccination in previously infected participants, particularly anti-SNURF IgG, but found no rise in thyroid peroxidase antibodies and did not establish a link between the detected antibodies and vaccination adverse effects. [6] However, prior infection(s) may have increase thyroid peroxidase antibodies sufficiently to yield no measured increase after vaccination, a type of positive control for the multiply vaccinated. Or, perhaps spike was not involved in the pathogenic priming again thyroid peroxidase.
The combination of past results helps frame the next question. Temporal evidence can establish that an abnormality appeared after an exposure; functional evidence can establish that an abnormality has biological effects. Connecting those findings within the same clinical phenotype would strengthen causal inference further. Combining observations across papers requires attention to whether they concern the same antibody, target and disease process.
Bodansky and colleagues’ 2024 study of multisystem inflammatory syndrome in children provided a more specific mechanistic connection. The discovery comparison included 199 children with MIS-C and 45 previously infected controls without MIS-C. Follow-up investigations linked immune recognition of SARS-CoV-2 nucleocapsid and the human protein SNX8, including cross-reactive T cells. This supplied direct support for molecular mimicry in a defined post-infectious syndrome. SNX8 was not an original candidate in my 2020 table. The relevant viral protein was nucleocapsid; spike-only vaccines do not encode it. [7]
While a mechanism demonstrated after infection alone cannot establish a vaccine-specific causal chain merely because both settings involve an immune response.This result strengthens the biological case for investigating molecular mimicry while preserving the need to demonstrate the relevant mechanism after vaccination. It also points out a key fact known in 2020: Exposure matters. Antigen identity matters. Spike or non-spike protein epitope exposure matters. Vaccine maker should have removed the unsafe epitopes to reduce the incidence of repeated exposure to auto-reactogenic epitopes and reduce pathogenic priming.
What the Yale PVS study contributes
The February 2025 version of the Yale LISTEN preprint, “Immunological and Antigenic Signatures Associated with Chronic Illnesses after COVID-19 Vaccination,” examined 42 participants with persistent symptoms categorized as post-vaccination syndrome and 22 vaccinated healthy controls. This was a selected, cross-sectional comparison without pre-vaccination blood samples from the affected participants. It could identify differences associated with membership in those groups. It could not determine how frequently vaccination produced the syndrome or establish when each immune difference began. [8]
Its supplementary antibody results deserve substantive attention. Table S4 reported 105 higher antigen–antibody-class measurements in the PVS group at the authors’ adjusted significance threshold: 65 IgM, 39 IgA and one IgG. It also reported 23 lower measurements, comprising 18 IgG and five IgA. These counts describe assay comparisons, not 105 distinct diseases, patients or newly induced autoantibodies. [8]
The reported higher IgA signals included aquaporin-4, thyroid peroxidase, liver cytosol antigen type 1 and gastric H+/K+-ATPase. Myelin basic protein signals moved in the opposite direction for IgA and IgG. The pattern therefore raises specific questions about neural, thyroid, hepatic and gastric antigen recognition while challenging any claim of a uniform increase in antibodies against nervous-system targets. [8] And some of these protein targets should, by now, sound familiar to you.
That degree of specificity improves the research question. It directs attention to antibody class, antigen and clinical phenotype. It validates prior findings yet again, thus helping put the spotlight on earlier predictions.
The smaller subgroup classified without evidence of prior infection did not reproduce every whole-cohort association at the adjusted threshold. This matters because while pathogenic priming does not care whether infection occurs first or injection, lower statistical power and exposure classification complicate interpretation [8]
The preprint advances the inquiry by documenting a clinically selected population and a set of biological differences. Its findings support targeted replication and longitudinal investigation. Establishing the contribution of vaccination, infection, pre-existing susceptibility and subsequent illness will require evidence that distinguishes those explanations.
Vaccine-specific observations deserve direct investigation
Semmler and colleagues compared 191 participants with chronic fatigue and dysautonomia after vaccination with 89 healthy controls. They reported differences in receptor-antibody measurements and inflammatory markers. The design included pre-vaccination and later samples for the healthy controls, while comparable pre-vaccination measurements were unavailable for the affected group. That distinction changes the interpretation: a higher measurement relative to healthy recipients after vaccination can reflect a different trajectory, including failure to show the decline observed in controls, without demonstrating newly induced antibodies in patients. The paper supplied a clinical-serological association that warrants replication and functional follow-up. [9]
The next study should resolve that temporal uncertainty. It should also determine whether the measured antibodies alter receptor function, track symptom severity or identify a subgroup with a reproducible biological course. Each result would answer a question the original comparison could not settle.
Thurner and colleagues’ investigation of myocarditis after vaccination added functional evidence concerning antibodies against interleukin-1 receptor antagonist, IL-1RA, an endogenous regulator of inflammation. Among biopsy-confirmed cases, the investigators detected these antibodies in nine of 12 patients younger than 21 and three of 28 older patients, compared with two of 214 vaccinated controls. Plasma from antibody-positive patients impaired IL-1RA activity in vitro. The proposed explanation involved an altered form of the human protein and loss of immune tolerance; the authors discussed prior evidence against cross-reactivity with viral structural proteins. Antibody-positive confirmed cases had a milder clinical course than antibody-negative cases. [10]
That finding expands the mechanisms requiring investigation. A prevention program focused exclusively on sequence mimicry could miss harmful responses arising through altered self-proteins or other disturbances of immune regulation. The appropriate response to a proposed mechanism is to test its predictions while remaining open to a better explanation.
It is a reminder the autoreactive antibody route to autoimmunity from pathogenic priming is not the only possible route.
Negative evidence belongs in the same account
A cumulative assessment must allow well-designed negative results to change the explanation.
Jaycox and colleagues studied 145 healthy vaccine recipients, 38 recipients with autoimmune disease and eight myocarditis cases. Their extracellular-antigen profiling found broadly stable IgG autoreactivity after mRNA vaccination, with follow-up extending to approximately four months in the study. This is meaningful evidence against a generalized increase in the extracellular IgG responses measured in those cohorts. The assay used purified IgG and did not assess IgA, IgM or intracellular targets. It therefore does not directly test the dominant antibody-class findings reported in Yale’s selected chronic PVS cohort. The eight myocarditis cases also did not reproduce the anti-IL-1RA finding. [11]
The result constrains a hypothesis where its measurements overlap that hypothesis. Its boundaries remain part of the interpretation. Unmeasured antibody classes cannot become negative findings through generalization, just as a positive result in a selected symptomatic group cannot establish risk in all recipients.
Patel and colleagues supplied another relevant negative result. They examined cardiac-tissue binding by IgG, IgM and IgA from ten patients with myocarditis after vaccination and ten with MIS-C, using tissue from two adult donor hearts. They did not detect binding above background in those groups. This study directly challenges explanations that predict readily detectable binding to the tested cardiac tissues. Its small cohorts and adult tissue substrate define the scope of that challenge. The experiment does not resolve every possible mechanism involving inflammation regulators, modified proteins or cellular immunity. [12]
These findings should shape subsequent studies. A research program that continues to propose the same detectable cardiac-binding mechanism must explain why it failed these tests or expand the proposal. Multiple mechanism of iatrogenic disease could exist. A program investigating a different target or antibody class or a different mechanism should state the expansion of explanations explicitly.
Or, it could be what Popper called a critical test of the hypothesis.
Either way, calling every negative study final or irrelevant would make the theory resistant to evidence. Treating every negative study as universal reassurance would make the safety assessment equally resistant to evidence. The solution is to specify the prediction, population, assay and observation period before drawing the conclusion.
Fatal vaccine injury has a documented immune example
The discussion of vaccine-induced death can rest on a concrete clinical example.
Greinacher and colleagues described 11 patients in Germany and Austria who developed an unusual syndrome of thrombocytopenia and thrombosis after the ChAdOx1 adenoviral-vector vaccine. Six died. In an expanded, selected laboratory series, all 28 specimens that screened positive for antibodies against platelet factor 4 produced PF4-dependent platelet activation. Additional antibody studies supported an immune mechanism. The work connected a distinctive clinical presentation with a measurable and functional abnormality. [13]
Pavord and colleagues subsequently characterized a separate UK series of 220 definite or probable cases of vaccine-induced immune thrombocytopenia and thrombosis, VITT, after a first ChAdOx1 dose. At analysis, 49 patients had died, 170 were alive and one outcome was unavailable. The reported mortality of approximately 22% described those selected VITT cases. It did not describe the probability that a vaccine recipient would die. [14]
Together, these investigations connected exposure, a recognizable syndrome, laboratory findings, functional evidence and fatal outcomes. The clinical series did not need to identify every initiating molecular event before the immune injury could become intelligible. The functional studies did not need to estimate population incidence to explain a crucial component of the disease process.
VITT demonstrates why cumulative inference matters. It also demonstrates why causal claims must retain their boundaries. This evidence concerns a specific syndrome associated with a particular vaccine platform. It does not validate every proposed vaccine injury, establish spike–human sequence mimicry as VITT’s initiating mechanism or provide a general estimate of vaccine-attributable mortality. [13,14]
An evidence base can establish that an intervention caused a particular fatal immune syndrome while leaving broader questions about other syndromes unresolved. Scientific precision requires retaining both conclusions.
Accumulation requires more than adding papers
The strength of a body of evidence depends on what its studies contribute and how independently they contribute it.
A review that summarizes an experiment adds interpretation, not new data. But it states contextual relevance. A secondary analysis of the same cohort can provide a valuable new endpoint while retaining the original cohort’s selection problems. Repeated use of the same assay can reproduce both a biological signal and an assay artifact. Publication count is therefore a poor substitute for an explicit account of participants, measurements and independent replications.
The central inferential question concerns competing explanations. A functional result that an antibody disrupts a relevant process may be substantially harder to explain as incidental background autoreactivity than a single positive binding measurement. A pre-exposure sample may distinguish newly appearing antibodies from a longstanding trait. A well-matched disease control may reveal whether a signal accompanies vaccination, persistent illness generally or a particular clinical syndrome.
Those contributions can reinforce one another. Their combined weight still depends on whether they concern the same causal pathway and whether shared errors explain their apparent agreement.
Statistical interpretation needs the same discipline. A p-value above 0.05 does not establish equivalence between groups. Sample size matters. Directionality of the test matters and it can be based on what an investigate expects – given prior evidence. A claim that a study excludes a clinically meaningful increase in risk requires adequate precision around the estimated effect. Conversely, a large screening panel creates opportunities for chance findings. A prior biological hypothesis can justify a prespecified test; it cannot retroactively exempt every favorable result from multiple-comparison scrutiny.
Study design also determines which harms can become visible. A shortened observation period can address an acute event while leaving delayed outcomes unresolved. An irrelevant surrogate endpoint can be positively or negatively misleading. A broad composite endpoint can obscure a change in one component. A comparison of symptomatic recipients with recipients selected because they remained healthy can identify a candidate biological distinction; estimating vaccine-attributable risk requires an appropriate population denominator and exposure comparison. These are properties of the question and design, not verdicts that can be inferred from a study’s conclusion.
Research into chronic illness must also account for people who decline further doses after becoming ill, differences in prior infection and diagnostic access, and losses to follow-up. The appropriate analytical treatment depends on when those factors arose and how they relate to exposure and outcome. Labeling every measured difference a confounder can be as misleading as ignoring confounding altogether.
Transparency allows these issues to be resolved. Investigators should make case definitions, exclusions, assay performance, transformations, analysis code and the full set of tested outcomes available to independent scrutiny. Financial and institutional interests belong in the interpretation of every study, regardless of whether its findings support or challenge a safety concern. Evidence of a discrepancy warrants an inquiry; an allegation of suppression or deliberate manipulation requires additional evidence of its own.
Prevention is the reason to do the work
The practical objective is to prevent avoidable harm while measuring the benefits and risks of medical interventions accurately.
The threshold for funding a targeted investigation should differ from the threshold for attributing an individual death or estimating a population death burden. A reproducible biological signal in a defined clinical population can justify further investigation before researchers have solved every causal question. Clinical care can address a patient’s documented impairments while the initiating mechanism remains under investigation. These decisions should state the uncertainty they can tolerate and the consequences of waiting.
For vaccine development, the pathogenic-priming question supports systematic evaluation of potentially harmful immune recognition alongside the evaluation of protective responses. Computational screening can prioritize questions. Evidence from human immunology in the earliest phases, clinical outcomes and independent replication in later phase of research must determine whether a proposed safety improvement reduces injury while preserving benefit. A design change should earn that conclusion through testing. This means more arms in the studies, but fewer casualties from vaccine injury. This means safer vaccines if conducted properly.
For persistent illness after vaccination, the immediate scientific obligation is to build studies that can distinguish the leading explanations. Prospective samples, carefully documented infections and exposures, clinically relevant outcomes and transparent analysis would make the next result more informative. The point of identifying an evidentiary gap is to design the study that can close it.
There are some who would prefer each new corroborative and validating study to be read as if it were the first study conducted on the planet on a topic because the logical error preserves their personal, professional, or financial prior bias. They want to impose the error of arguing from ignorance on the scientific and medical community. This happens more often than most know. “That result is merely theoretical” make no sense in the handling of evidence in a Popperian manner. “Given background knowledge, that result corroborates (or falsifies) the existing hypothesis” makes a great deal of sense and prevents narrative enforcement in the name of science.
The record has advanced since April 2020. It now includes experimental tissue recognition, longitudinal observations of autoantibodies after infection, functional immune abnormalities, direct molecular-mimicry evidence in a post-infectious syndrome, selected studies of illness after vaccination and a documented vaccine-induced immune syndrome with fatal outcomes. The findings support different conclusions at different levels of specificity. Their scientific value grows when those relationships are examined explicitly. [1–14]
No study is an island. Each credible result enters an existing body of knowledge and changes what the next investigation should ask. We should expect and demand accurate attribution for prior predictions, serious tests of proposed mechanisms, full consideration of contrary results and clinical accountability for preventable harm.
The standard is demanding because the purpose is consequential: identify the people at risk, understand the injury, improve the intervention and prevent the next avoidable case.
References
[1] Lyons-Weiler J. Pathogenic priming likely contributes to serious and critical illness and mortality in COVID-19 via autoimmunity. Journal of Translational Autoimmunity. 2020;3:100051. Published April 9, 2020. DOI: 10.1016/j.jtauto.2020.100051. PMID: 32292901. Primary article.
[2] Vojdani A, Kharrazian D. Potential antigenic cross-reactivity between SARS-CoV-2 and human tissue with a possible link to an increase in autoimmune diseases. Clinical Immunology. 2020;217:108480. DOI: 10.1016/j.clim.2020.108480. PMID: 32461193. Primary article.
[3] Vojdani A, Vojdani E, Kharrazian D. Reaction of Human Monoclonal Antibodies to SARS-CoV-2 Proteins With Tissue Antigens: Implications for Autoimmune Diseases. Frontiers in Immunology. 2021;11:617089. Published January 19, 2021. DOI: 10.3389/fimmu.2020.617089. PMID: 33584709. Primary article.
[4] Timofeeva AM, Aulova KS, Mustaev EA, Nevinsky GA. SARS-CoV-2 Spike Protein and Molecular Mimicry: An Immunoinformatic Screen for Cross-Reactive Autoantigen Candidates. International Journal of Molecular Sciences. 2025;26(18):8793. DOI: 10.3390/ijms26188793. PMID: 41009363. Primary article.
[5] Wang EY, et al. Diverse functional autoantibodies in patients with COVID-19. Nature. 2021;595:283–288. DOI: 10.1038/s41586-021-03631-y. PMID: 34010947. Primary article.
[6] Jernbom AF, et al. Prevalent and persistent new-onset autoantibodies in mild to severe COVID-19. Nature Communications. 2024;15:8941. DOI: 10.1038/s41467-024-53356-5. PMID: 39414823. Primary article.
[7] Bodansky A, et al. Molecular mimicry in multisystem inflammatory syndrome in children. Nature. 2024;632:622–629. DOI: 10.1038/s41586-024-07722-4. PMID: 39112696. Published article, author repository.
[8] Bhattacharjee B, et al. Immunological and Antigenic Signatures Associated with Chronic Illnesses after COVID-19 Vaccination. medRxiv. 2025.02.18.25322379. Version 1, posted February 18, 2025; preprint. DOI: 10.1101/2025.02.18.25322379. Analysis here refers to that version and its supplied supplementary figures and Tables S1–S8. Version-specific record.
[9] Semmler A, et al. Chronic Fatigue and Dysautonomia following COVID-19 Vaccination Is Distinguished from Normal Vaccination Response by Altered Blood Markers. Vaccines. 2023;11(11):1642. DOI: 10.3390/vaccines11111642. PMID: 38005974. Primary article.
[10] Thurner L, et al. IL-1RA Antibodies in Myocarditis after SARS-CoV-2 Vaccination. New England Journal of Medicine. 2022;387:1524–1527. DOI: 10.1056/NEJMc2205667. PMID: 36130012. Primary article.
[11] Jaycox JR, et al. SARS-CoV-2 mRNA vaccines decouple anti-viral immunity from humoral autoimmunity. Nature Communications. 2023;14:1299. DOI: 10.1038/s41467-023-36686-8. PMID: 36894554. Primary article.
[12] Patel H, et al. Evaluation of Autoantibody Binding to Cardiac Tissue in Multisystem Inflammatory Syndrome in Children and COVID-19 Vaccination-Induced Myocarditis. JAMA Network Open. 2023;6(5):e2314291. DOI: 10.1001/jamanetworkopen.2023.14291. PMID: 37200028. Primary article.
[13] Greinacher A, et al. Thrombotic Thrombocytopenia after ChAdOx1 nCov-19 Vaccination. New England Journal of Medicine. 2021;384:2092–2101. DOI: 10.1056/NEJMoa2104840. PMID: 33835769. Primary article.
[14] Pavord S, et al. Clinical Features of Vaccine-Induced Immune Thrombocytopenia and Thrombosis. New England Journal of Medicine. 2021;385:1680–1689. DOI: 10.1056/NEJMoa2109908. PMID: 34379914. Published article, university repository.
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
Leave a Reply