INTERACTIVE MODELFIELD NOTEPALEOGENOMICS

The Diagnostic Position Bench

A live model accompanying “The Denisovans Finally Have a Face”

Study: Field note
Review: Peer-edited by the club review board
Article: 13 sectionsabout 20 min
Files: LaTeX source30 kB

The Harbin identification rests on two things: a small set of mitochondrial positions that only Denisovans carry, and the fact that a specimen with an 85-year gap in its custody chain has to survive a hard question about contamination. Below, mix contamination into a sample yourself and watch what survives. Then drag a marker across a million years of mitochondrial history and see where the skull actually sits.

27 diagnostic positions / 16,569 bp mtDNA ≈ 0.16% of the sequence; 1,000,000 ya ÷ 660,000 ya ≈ 1.5×

Model 1. Mix in some contamination

A simplified mitochondrial panel below shows 27 positions where every previously sequenced Denisovan genome agrees and modern humans do not — the real diagnostic set described in the paper, just drawn much denser than its true 1-in-614 spacing across the genome so you can see every position at once. Slide the contamination level and each position is redrawn: with probability equal to the contamination level, a modern human read wins that position instead of the true ancient one. Positions that are not diagnostic are shown too, dimmed, to make the point that contaminating them teaches you nothing either way.

Contamination level20%

Draw a new sample

diagnostic positions read: 27still showing ancient allele: 22showing modern contamination: 5false Denisovan calls from contamination: 0

Even at moderate contamination, the positions that still show the ancient allele are still Denisovan-consistent, and modern contamination cannot generate a false one, because a contaminating read is by definition a modern human allele, not a Denisovan one.

This is the club's own illustrative model, not the paper's statistical method. It is built to show one real thing: contamination can only ever erase a diagnostic signal by replacing it with a modern human read, never manufacture a Denisovan one. That asymmetry, not a specific contamination threshold, is what the 27-position argument in the field note relies on.

Model 2. A million years of mitochondrial history

Drag the marker below across time. It moves along whichever lineage you have selected, and the readout tells you where it sits relative to the two real branch points from the primary sources: the Neanderthal–modern human mtDNA split at about 660,000 years ago, and the Denisovan mtDNA split from that combined lineage at about 1,000,000 years ago. The shaded band marks the Harbin skull's own independently dated age, 146,000 to 309,000 years old.

Years before present200,000LineageDenisovan

Marker sits at: 200,000 years ago, on the Denisovan branch
Relative to Denisovan split: 800,000 years after the ~1,000,000 ya split
Relative to Harbin skull: outside the skull's dated bracket

Branch points from Krause et al. 2010 (~1,000,000 ya) and Green et al. 2008 (660,000 ± 140,000 ya). Harbin's age bracket, 146,000–309,000 years, comes from Shao et al. 2021's independent direct dating of the specimen, not from the mtDNA itself.

Notice how little of the Denisovan branch's length the Harbin skull actually occupies. The lineage itself is old; the individual whose plaque we can now read is comparatively recent, sitting well after the split, in roughly the same span of time several other Denisovan-associated fossils across Asia also date to.