The first Denisovan forearm bone came out of a 22-sample shortlist

The arithmetic behind the triage is the part other sites can copy. Sending 22 fragments to the mass spectrometer rather than the full assemblage is roughly 2,900 times fewer runs ^[3], and three of the 22 came back hominin, a hit rate near 14 percent ^[1]. Blind ZooMS screening at Denisova Cave and Baishiya Karst Cave has recovered hominin bone at about 0.1 percent of specimens ^[14], so on those two numbers the pre-screen improved yield per sample by something like a hundredfold ^[2]. The comparison flatters the new method, though: 0.1 percent is a site-level recovery rate across everything screened, while 14 percent is a hit rate inside a subset chosen precisely because it already looked like hominin bone ^[5].

Fragments that did not look diagnostic were never tested. Three hominin bones is therefore a floor on what the assemblage holds rather than a measurement of it, and nothing in the design constrains how many hominin pieces stayed in the discard trays.

Before this excavation, everything known of the Denisovan body below the skull was a finger bone and a rib fragment ^[15]. The Bianfu radius is a proximal piece, and its head, where the forearm meets the humerus at the elbow, is closer to Neanderthal form, while the bone's overall shape sits nearer modern humans ^[16]. Fu told Live Science that the human-like form "likely provided functional advantages for lightweight, precise manipulation" ^[17]. One proximal fragment from one individual: the shape is data, the function is inference, and there is no population range around it yet.

Each specimen gave between six and 16 endogenous proteins, covering 1,972 to 4,076 amino acid residues ^[9], and Bayesian trees built from those consensus sequences reproduced the topology of nuclear-genome studies, with every Bianfu sample falling into a clade with Denisova 3 ^[20].

The dated fossils span about 33,000 years inside a cultural sequence of about 120,000, near 28 percent of the occupation ^[4], and the Nature paper describes that record as the one most parsimoniously attributed to Denisovans ^[19]. The layers with no hominin bone in them are assigned by continuity of technology, not by protein.

Much of the behavioural evidence is independent of the molecular work. There are 1,654 identifiable animal bones, dominated by deer and bovids, with rhinoceros, hyena and bamboo rat also present; many carry butchery marks and few show non-human gnawing ^[22]. The stone kit runs to 1,366 artefacts of flakes, cores, scrapers and choppers, simple next to contemporaneous Neanderthal assemblages ^[21], used alongside pervasive unmodified bone in a conifer-dominated forest or forest-steppe ^[23]. Bence Viola of the University of Toronto, who was not involved in the work, calls Bianfu only the third Denisovan site "where we have everything," after Denisova Cave and Baishiya Karst Cave ^[24]. That sets the size of the comparative sample any of this feeds: two sites with full context before, three now ^[5].