Bronze Age England: Where Do the Closest Genomes Lie?

Where do the closest genomes to Bronze Age England lie? Three methods are applied to the same 20 Beaker-period persons (2380–1990 BCE) and 25 Middle Bronze Age persons (1480–1130 BCE) from England. A Gaussian-process match map following the principle of Mobest¹ places the closest contemporaneous genomes of the Beaker-period persons on the continent (Germany 13 %, Sweden 10 %, France 10 % of the land probability) and the best match of none of the 20 persons inside Britain and Ireland; the map of the Middle Bronze Age persons is flatter (Germany 12 %, Britain and Ireland 11 %, France 10 %). In entropy-rescaled PCA distances² the Beaker-period persons lie closer to the Netherlands Beaker group than to the English Neolithic in every panel (Q = 0.07, 95 % interval 0.02–0.14; Czechia Beaker 0.19, 0.14–0.23). In qpAdm, 93.8 ± 1.3 % Netherlands Beaker and 6.2 % English Neolithic fit the Beaker-period persons on right panel D (p = 0.69) and are rejected on right panel P (p = 0.0018), where the weights are the same; the English Neolithic alone is rejected on both. For the Middle Bronze Age persons the English Beaker group alone is rejected (p < 10⁻⁹), and two models with an added source, English Neolithic (9–10 %) or French Early Bronze Age (23–26 %), are retained on both panels. Positions on a map are not ancestry.

A Beaker-associated turnover in Britain was reported by Olalde et al.³ This article asks where the closest genomes to two groups of persons from England lie, with three methods that read different things: where on a map the contemporaneous genomes most similar to a person lie (a match map¹), which of two groups a person is closer to on an entropy-rescaled PCA map², and which mixtures of sampled source groups fit the persons in qpAdm. The 20 Beaker-period and 25 Middle Bronze Age persons are the persons of the Gaussian-process maps of England, so one set of persons is read three times. A map position and a PCA distance are not ancestry; only the qpAdm models give weights, and these hold for the stated sources and right panels.

Persons. The persons are those of the Gaussian-process match maps of England: 20 Beaker-period persons (2380–1990 BCE; group GP_England_BellBeaker, called England Beaker below) and 25 Middle Bronze Age persons (1480–1130 BCE; GP_England_MBA). The same persons enter all three methods.

Match maps. Own Gaussian-process implementation following the principle of Mobest¹ (not the Mobest software): a regression over space and time (kernel exp[−(Δs²/800² + Δt²/800²)], 800 km and 800 years) of four Nexju N25 PCA coordinates over 12,990 ancient individuals of the AADR v66. Each person is placed at their own date, the persons shown are left out of the reference, and the map of a group is the mean of the maps of its persons. Shares by country refer to land only; the share of persons whose cell of highest probability lies inside Britain and Ireland and the share of the probability on sea cells are reported with them. Peaks follow where DNA has been sampled, not necessarily where people lived.

Rescaled PCA distances. Entropy-rescaled PCA distances after Taleb et al.² (maps A and B) in the Nexju PCA-distance series, over three reference panels (WE8 with shrinkmode projection, WE8 with plain projection, EU5 with shrinkmode projection), that is six settings, and 16 shifts of the origin. Q is the distance of a target to the first group of a contrast over its distance to the second; below 1 the target is on the side of the first group. Intervals are 95 % bootstrap intervals (2,000 resamples of samples within groups) on the base panel, WE8 with shrinkmode.

qpAdm. qpAdm on the Nexju platform with alpha 0.05 and allSnps = false, on two right-group panels fixed before the runs: panel D (primary; 14 right groups; 0.70–0.72 M SNPs) and panel P (regional; 11 right groups; 1.00–1.04 M SNPs). A model is retained when p ≥ 0.05, all weights lie between 0 and 100 % and the weight SE is below 10 percentage points; otherwise it is rejected. The models were declared before the runs: for the Beaker-period persons Netherlands Beaker + England Neolithic (BB1), Czechia Beaker + England Neolithic (BB2) and each of Netherlands Beaker (BB3) and England Neolithic (BB4) alone; for the Middle Bronze Age persons England Beaker alone (MB1), England Beaker + England Neolithic (MB2) and England Beaker + France Early Bronze Age (MB3). The right-panel sensitivity of BB1 was run from each panel as base, dropping one right group at a time and swapping the panel.

Match maps. For the 20 Beaker-period persons the five largest shares of the land probability are Germany 13 %, Sweden 10 %, France 10 %, Russia 9 % and Ukraine 9 % (Table 1). Britain and Ireland are not among them, none of the 20 persons has its cell of highest probability inside Britain and Ireland (0 %), and 35 % of the probability of the map falls on sea cells. The 25 Middle Bronze Age persons give a flatter picture: Germany 12 %, Britain and Ireland 11 %, France 10 %, Sweden 8 % and Norway 7 %; 3 of 25 persons (12 %) have their best match inside Britain and Ireland, and 44 % of the probability falls on sea cells. The Beaker-period result is consistent with the Beaker-associated turnover reported by Olalde et al.³

Rescaled PCA distances. The Beaker-period persons lie closer to the Netherlands Beaker group than to the English Neolithic: Q = 0.07 (95 % interval 0.02–0.14) on map A and 0.07 (0.02–0.14) on map B, and this side holds in all six settings and all 16 origin shifts (Fig. 1, Table 2). Against Czechia Beaker, Q is 0.19 (0.14–0.23) on map A and 0.18 (0.14–0.23) on map B; against Germany Beaker it is 0.23 and 0.22. All four contrasts of the series put the Beaker-period persons on the side of the Beaker group, with intervals that exclude 1. The Middle Bronze Age persons also lie on the side of the Beaker groups in all four contrasts (Q from 0.12 to 0.23).

qpAdm. For the Beaker-period persons, 93.8 ± 1.3 % Netherlands Beaker and 6.2 % England Neolithic fit on right panel D (p = 0.687, χ² 9.2 with 12 df; Table 3). On panel P the weights are nearly the same (93.1 % and 6.9 %, SE 1.3 pp) and the model is rejected (p = 0.0018). England Neolithic alone is rejected on both panels (p ≈ 0), as are Netherlands Beaker alone (p = 0.0046 on panel D) and Czechia Beaker with England Neolithic (p = 0.021 on panel D, 0.0003 on panel P). On panel D all 13 variants that drop one right group fit (p from 0.48 to 0.87; Netherlands Beaker 93.6–94.3 %; Fig. 2). On panel P, 9 of the 10 such variants are rejected and only the variant without Sweden_Rossberga_N fits (p = 0.105; Fig. 3); the model counts as rejected on panel P.

For the Middle Bronze Age persons, England Beaker alone is rejected on both panels (p = 2.3×10⁻¹⁰ and ≈ 3×10⁻¹⁵; Table 4). England Beaker with England Neolithic (90–91 % and 9–10 %) and England Beaker with France Early Bronze Age (74–77 % and 23–26 %) are retained on both panels (p from 0.079 to 0.334). The runs do not separate the two.

Beaker period. The three methods point the same way. The map places the closest contemporaneous genomes on the continent; in rescaled PCA distances the persons lie closer to the Netherlands Beaker group than to the English Neolithic in all six settings; in qpAdm Netherlands Beaker with 6 % English Neolithic fits on panel D, and England Neolithic alone is rejected on both panels. qpAdm fits on one of the two right-panel sets only.

Middle Bronze Age. The English Beaker group alone does not fit. In both retained models it carries 74–91 % of the persons, and the rest is English Neolithic (9–10 %) or French Early Bronze Age (23–26 %); these runs do not tell the two apart. The map of the Middle Bronze Age persons is flatter than that of the Beaker-period persons, with Britain and Ireland second (11 %).

What the methods show and what they do not. A map position and a PCA distance describe where similar genomes lie, not ancestry, and peaks follow where DNA has been sampled. Only qpAdm gives weights, and these hold for the stated sources and right panels. A source in a model is a sampled group; its name does not say where the ancestors lived.

Data and methods. Match maps: own Gaussian-process implementation following the principle of Mobest (Schmid & Schiffels 2023); not the Mobest software, not PANE, not qpAdm; exploratory, post hoc. Training data: 12,990 ancient individuals of the AADR v66 (29.9–70 °N, −24–70 °E, 150–8500 BP, at least 15,000 SNPs, one version per person, training window ±2000 years); response: four Nexju N25 PCA coordinates (European PC1–2, Middle-Eastern PC1–2); kernel 800 km and 800 years; nugget 0.4 and the variance of each coordinate fitted by marginal likelihood on a random subset of 5,008 of the 9,271 reference individuals dated 1,080–6,328 BP; 5,124–8,226 training individuals per person; map: product of normal densities, normalised to 1, 40 km grid without land mask, smoothed to 10 km by cubic interpolation; the 45 persons shown are left out of the reference. Validation: on Albania, a Balkan peak appeared for 2 of 4 test persons. The maps were computed with the author's own code; they are not a Nexju analysis run. Rescaled PCA distances: Nexju PCA-distance run on maps A and B with the base panel WE8-shrinkmode (992,097 SNPs) and the settings WE8-plain and EU5-shrinkmode (970,217 SNPs); 16 origin shifts; bootstrap of 2,000 resamples, 95 %; a statement is stable when it holds in at least 5 of 6 settings, at least 13 of 16 shifts and in the bootstrap. qpAdm: Nexju qpAdm, alpha 0.05, allSnps = false. Panel D (14 right groups): Australian, Papuan, Mbuti, Karitiana, Turkey_PPN, Israel_Natufian, Iran_WezmehCave_N, Italy_Sicily_Epigravettian, Turkey_N, Russia_Samara_EBA_Yamnaya, Iran_GanjDareh_N, Georgia_KotiasKlde_Mesolithic, Serbia_IronGates_Mesolithic, Russia_Karelia_Mesolithic. Panel P (11 right groups): Australian, Papuan, Russia_Samara_EBA_Yamnaya, Iran_GanjDareh_N, Georgia_KotiasKlde_Mesolithic, Sweden_Mesolithic_HG, Germany_GrossFredenwalde_Mesolithic, France_Yonne_N, Austria_N_LBK, Germany_BlatterhohleCave_MN, Sweden_Rossberga_N. Sources: Netherlands_LNB_BellBeaker, Czechia_BellBeaker, England_N, France_EBA and the 20 Beaker-period persons (GP_England_BellBeaker). Targets: GP_England_BellBeaker (20) and GP_England_MBA (25). Models and right-group panels were fixed before the runs.

Magnus Albertus Papers · @AlbertusMvagnus