From molecular detection to historical diagnosis: interpretative challenges in paleoproteomics

1 A broad field, and a narrow claim

Paleoproteomics has grown from a handful of studies on individual abundant proteins into a discipline whose applications now range from the taxonomic identification of highly fragmented bones and shells and the phylogenetic resolution of extinct species, through biological sex estimation based on sexually dimorphic amelogenin peptides preserved in dental enamel, to the exploration of past cuisines from dental calculus and pottery food crusts and the characterization of past diseases; more broadly still, it has opened onto past human–animal interactions, the reconstruction of past environments and environmental change, and the expansion of the hominin fossil record through large-scale screening of nondiagnostic bone fragments (Parker et al., 2019; Stewart et al., 2017; Warinner et al., 2022). The recovery of Early Pleistocene enamel proteome sequences resolving Stephanorhinus phylogeny illustrates how far beyond the reach of ancient DNA these methods now extend (Cappellini et al., 2019).

This breadth matters, because the difficulty examined here is not general to the field. It is specific to one class of claim, and the distinction turns on what the recovered molecule is asked to establish.

When a peptide identifies a bone fragment as ovicaprine, or a milk protein in dental calculus indicates dairy consumption (Warinner et al., 2022), detection establishes presence, and presence is very nearly the whole of the claim; the residual questions - contamination, database bias, differential preservation between taxa - are technical, acknowledged, and increasingly well controlled. When a microorganism recovered from the remains of a named individual is proposed as the cause of that individual's death, detection is asked to establish something categorically different: not that the organism was there, but that it was responsible. Between presence and responsibility lies a gap that no gain in analytical sensitivity can close, because the gap is not analytical.

2 When detection approximates diagnosis, and when it does not

Clinical medicine enforces this distinction by daily necessity. Recovering an organism from a patient establishes that the organism is present; whether it caused the illness is a separate judgement resting on the chronology of symptoms, the anatomical distribution of lesions, the tempo of the disease, the epidemiological setting, and the plausibility of the proposed mechanism. Blood-culture contaminants, colonising flora, and organisms from non-sterile sites are all detected without being diagnostic.

There are nonetheless circumstances in which the two converge, and specifying them defines the boundary. When an organism is obligately pathogenic, is recovered from a normally sterile compartment, and is accompanied by a clinical picture that matches its known behaviour, the inferential distance collapses. What licenses the shortcut is the conjunction of the three conditions - not the sensitivity of the assay. The third is the one most often neglected in retrospective work, and it is the one that does the most work: an organism recovered from a protected compartment still requires that the illness it is said to have caused resemble the illness it is known to cause.

3 Two ancient biomolecules, two authentication traditions

Proteins persist under conditions and across timescales that exceed those of DNA by orders of magnitude, and this is precisely their value: they open windows paleogenomics cannot (Hendy et al., 2018). But the two fields differ in a respect bearing directly on causal claims.

Ancient DNA built its authentication apparatus in response to a sustained crisis of spurious results, and that apparatus - damage patterns as authenticity criteria, deamination signatures, independent replication, systematic blanks, sequence-level contamination screening - is now a precondition of publication. Paleoproteomics has no equivalent consolidated framework: there is at present no explicit consensus on standards for data reporting, validation measures, or suitable contamination controls, and no agreed criteria by which researchers, reviewers and editors may evaluate ancient protein data (Hendy et al., 2018). Contaminating proteins enter from the burial environment, from excavation, from storage and curatorial handling, and from the laboratory itself (Hendy et al., 2018).

Neither field is disqualified by this. But a diagnostic claim resting on ancient proteins currently carries less external authentication than the same claim resting on ancient DNA, which argues for restraint proportionate to the gap.

4 Two recent cases

Two studies published within months of each other make the problem concrete. Both are technically accomplished; neither is careless; and in both the difficulty lies at the same point.

Louis XIV. Paleoproteomic analysis of the king's embalmed heart yielded Cyphellophora europaea, and chromoblastomycosis, a chronic fungal infection primarily affecting the skin and subcutaneous tissues, was advanced as a possible cause of the death of 1715 (Charlier et al., 2026b). The identification is not in question. Its clinical interpretation is. The king's terminal course was marked by severe lower-limb pain followed by death within roughly twenty days, and the autopsy describes rapidly progressive extensive gangrene of the soft tissues with abdominal distension - a picture far more suggestive of an acute necrotizing septic process than of a chronic melanized fungal infection. Chromoblastomycosis is classically slowly progressive, confined to skin and subcutaneous tissue, and rarely disseminated. An acute bacterial necrotizing infection complicating a diabetic foot lesion, potentially with clostridial myonecrosis, accounts far better for the documented evidence (Demas, 2026a, b).

Raphael. Proteomic analysis of bone fragments attributed to the painter yielded a signature for Nocardia niigatensis, detected only in marrow from a rib fragment, and Nocardia was proposed as a possible cause of the death of 1520 (Charlier et al., 2026a). Nocardia species are environmental, soil-dwelling bacteria capable of causing opportunistic infections, most commonly involving the lungs. Here too the analysis is careful: the authors note that other, environmental bacteria were recovered from grave soil, that no sign of embalming was found, and - with commendable candour - that no conclusive evidence establishes the authenticity of the remains. Yet Nocardia is a ubiquitous soil saprophyte, the remains come from a burial exposed to flooding, and the environmental organisms recovered from the same site are precisely what makes exclusivity in a single marrow sample an uncertain discriminator. Meanwhile an alternative hypothesis fits the documented history at least as well: acute pulmonary histoplasmosis, an inhalational fungal infection acquired through exposure to airborne Histoplasma spores, contracted during the artist's well-attested descents into the Domus Aurea, whose enclosed subterranean chambers constitute a classical exposure setting (Demas, 2026a, b).

5 What the two cases share

Three features recur, and together they define the problem more precisely than any general appeal to caution.

First, in each case the organism was recovered and the diagnosis proposed, but the illness the organism is known to produce was not systematically compared with the illness the sources describe. Detection preceded clinical reasoning rather than being tested by it.

Second, in each case the recovered organism has an environmental reservoir - a soil saprophyte, an environmental fungus - which is the situation in which post-mortem origin is hardest to exclude, and in which the protected-compartment argument bears the greatest weight it can bear.

Third, and most importantly, in each case an alternative etiology remains compatible with the same historical record and was not searched for. This is the crux. Paleoproteomic identification is constrained by what is present in the interrogated databases and by what the analysis is configured to seek. Recovering one organism does not close a differential; it documents one presence within one compartment. Where a competing hypothesis accounts for the clinical course better and was never tested, the recovered organism cannot be treated as having resolved the question.

6 The gap between hedged claim and public conclusion

It should be said plainly that neither study asserts certainty. Both use “possible” in the title, and the authors of the Raphael study set out their own reservations at length. This is proper practice and deserves acknowledgement.

But the hedge does not survive dissemination. The Louis XIV study was reported publicly as having established that the accepted cause of death had been wrong for three centuries. A qualifier occupying one word of a title cannot restrain a conclusion that occupies a headline, and it does not restrain the citation practices of subsequent literature either. The problem is not authorial overreach. It is that lexical caution is not a substitute for a graded framework of causal attribution: where no such framework exists, a hypothesis and a finding are transmitted with equal force.

7 Toward a usable criterion

The two cases converge on a criterion, and it can be stated without appeal to any general principle of caution.

A paleoproteomic finding supports the attribution of a cause of death to a named individual to the extent that three conditions hold. First, the organism's known clinical behaviour must match the documented course in tempo and anatomical distribution - the condition that fails when a chronically progressive cutaneous mycosis is invoked to explain a twenty-day septic gangrene. Second, the organism's environmental distribution must make post-mortem acquisition implausible rather than merely unlikely — the condition that remains unsatisfied when a ubiquitous soil saprophyte is recovered from remains exposed to flooding, however protected the sampled compartment. Third, competing etiologies compatible with the same historical record must have been explicitly sought and excluded - the condition that neither study attempted.

The three are not equally weighted. The first governs the others, because an organism whose behaviour cannot produce the documented illness is not rendered causal by any degree of analytical rigour elsewhere. Where it fails, the finding should prompt reconsideration of what has been identified, rather than revision of what is known to have happened.

Nothing in this criterion is peculiar to ancient material. It is the ordinary structure of clinical attribution, applied to a setting in which the patient cannot be examined and the differential cannot be narrowed by any further investigation. That the setting is harder is an argument for applying the criterion more strictly, not less.

8 Conclusion

The interpretative burden carried by paleoproteomic findings varies with what they are asked to establish, and standards should vary with it. For taxonomic identification, dietary reconstruction, and phylogenetic placement, the field's methods are proportionate to its claims, and its results have transformed what can be known about the past. For attributing a cause of death to a named historical individual, they are not yet proportionate - and the deficit is not technological but inferential.

Paleoproteomics has opened a genuine window onto the biological past, and its sensitivity will keep increasing. Increasing sensitivity, however, expands the set of things that can be detected faster than it expands the set of things that can be concluded. In ancient biomolecular inference, the greatest risk is not failing to detect the trace, but mistaking detection for diagnosis.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Bethany Turner-Livermore, Georgia State University, United States

Reviewed by: Daniel Pinero, National Autonomous University of Mexico, Mexico