BGPT: Paper Review: A late Middle Pleistocene Denisovan mandible from the Tibetan Plateau

Explore by Goal

Quick Explanation Copied

BGPT take (skeptical, evidence-first)

Xiahe mandible is dated to a minimum of ~160 kyr via U-series on adhering carbonate crust, making it much earlier than known Tibetan Plateau occupations attributed to modern humans.
Denisovan attribution is based on palaeoproteomics: the dentine contains endogenous ancient proteins whose phylogeny places Xiahe with Denisovans (relative to great ape reference proteins), while ancient DNA signals are not detected in the reported mitochondrial capture.
Main risk to confidence: single-specimen evidence + phylogenetic inference from a limited set of informative proteins; morphology could also overlap with other Middle Pleistocene groups due to variability/convergence.

If you want, I can also run an “evidence-merging” critique plan that lists exactly what evidence would most easily disprove the Denisovan attribution.

Long Explanation

Paper Review (skeptical, visual-first): A late Middle Pleistocene Denisovan mandible from the Tibetan Plateau

Core claim: Xiahe mandible (~minimum 160 kyr) outside Denisova Cave is Denisovan/Denisovan-related, implying deep occupation and (indirectly) early selection for high-altitude adaptation in the Tibetan Plateau lineage.

1) Evidence map (what supports what?)

Evidence nodes

Geochronology: U-Th ages on adhering carbonate crust.
Ancient protein preservation: endogenous dentine proteins found; bone fraction negative in MALDI.
Proteomic phylogeny: recovered collagen protein sequences place Xiahe with Denisovan D3 (Bayesian probability reported as 0.99).
Ancient DNA: no evidence for ancient human mtDNA damage patterns under their capture/mapping criteria.
Morphology: robust archaic mandible traits + unusually large dental arcade; EDJ and arcade geometric morphometrics.

2) Visual geochronology (U-Th on adhering crust)

Interpretation (skeptical): The paper argues subsample agreement implies crust formed in a short interval around ~160 kyr and thus a minimum age for the mandible.

3) Timeline context (as stated in the paper)

Important limitation: the chart mixes a minimum fossil age with cited occupation dates and a broad MIS reference, so it is not a formal statistical posterior. The visual is only faithful to the paper’s stated directional comparisons.

4) Palaeoproteomics: what is the identification pipeline?

Key evidential strength: they explicitly report endogenous-protein indicators (deamidation, cleavage patterns), blank handling, and that bone returned negative MALDI spectra, which reduces (not eliminates) contamination confounds.

5) Negative ancient DNA result: what it does and doesn’t prove

Core observation

The authors report no ancient damage pattern in captured mitochondrial DNA, concluding the mandible lacks evidence for ancient human DNA.

What this means for confidence:

Supports that protein evidence is the main molecular handle here.
But does not rule out contamination-free ancient DNA being too degraded or absent in the sampled fraction; so “no ancient DNA” ≠ “no ancient DNA existed.” (This is a general interpretive caution consistent with their reported negative result.)

6) Morphology: where it’s informative vs where it can mislead

Skeptical caution: Middle Pleistocene mandible/dental morphospace can include substantial overlap; the paper itself discusses how Xiahe’s mandibular and dental arcade patterns differ in where they fall relative to groups (e.g., nearer H. erectus in mandibular shape vs away in dental arcade shape), which is informative but also highlights dependence on which traits/analyses are emphasized.

7) How strong is the Denisovan assignment?

Confidence-weighted breakdown

Chronology supports “late Middle Pleistocene” using consistent U-Th ages on adhering crust: strong for minimum age.
Proteomic phylogeny is the main molecular anchor (endogenous proteins + phylogenetic placement): strong, but still based on limited proteins and a single specimen.
Ancient DNA is negative, so genomic-level convergence checks are unavailable: reduces ability to adjudicate fine relationships.
Morphology supports “archaic Middle Pleistocene” and provides additional trait comparisons but can be less decisive for taxon assignment: moderate.

8) Wider evolutionary framing (where it fits)

The paper’s logic connects deep Denisovan presence to later high-altitude adaptations in Tibetans, but the paper’s molecular fossil data are indirect evidence for later allelic contributions. For broader context on how introgression complicates strict species-tree interpretations, see: .

9) What would most easily disprove the main conclusion?

Proteomic reanalysis that fails to reproduce endogenous deamidation/cleavage signals under stricter contamination controls or yields a different phylogenetic placement (protein orthology/filters matter).
Discovery of additional fossils from the region with contradictory proteomic phylogenetic placements or incompatible ages.
Independent dating mismatch: if new U-Th/other chronometers on directly associated materials yield substantially younger ages, the Denisovan attribution’s temporal context weakens.

Paper review metrics (BGPT scoring grid)

Scientific quality: 9/10 — multi-modal evidence (U-Th + CT-based morphology + proteomics) with explicit blank controls and use of phylogenetic methods.
Reproducibility: 9/10 — proteomics raw data accession PXD011377 and a public surface scan model are reported.
Main reproducibility risk: proteomic endogenous classification and protein phylogeny depend on pipeline choices (filters, contamination lists, deamidation/cleavage metrics), so fully independent replication could shift placement.

Optional BGPT actions (click to go deeper with focused, evidence-first prompts):

Run an independent “AI Scientist” analysis (iterative computation)

Author Reviews (BGPT)

Feedback:

Updated: March 22, 2026