Fig. 2: Ancestry from the Iron Age to the early medieval period in Europe.

a, Source groups used for qpAdm modelling of early medieval Europe. MDS is computed jointly with individuals from later periods using pairwise outgroup f₃ statistics (outgroup: Han Chinese people). These are calculated using Twigstats on Relate genealogies with a cut-off of 1,000 generations. The geographical map shows sampling locations of these individuals. b, The genetic structure of ancient groups predominantly from early medieval contexts shown on the same MDS as in a. The magnified inset shows an MDS computed without Twigstats on the same samples as the Twigstats MDS and focusing on early medieval or later individuals. c, Ancestry models of early medieval (EM) groups across Europe computed using qpAdm. Sample sizes are shown in black boxes. Sources are highlighted in a and marked as bold in the key, and were used in a rotational qpAdm scheme. For each target group, we remove models with infeasible admixture proportions (falling outside [0, 1]) and use a Twigstats cut-off of 1,000 generations. All models satisfy P > 0.01, unless a −log₁₀[P value] is shown next to the model. If models satisfy P > 0.05, we show all such models; otherwise, we show only the model with the largest P value. d, The ancestry proportion derived from EIA Scandinavia in groups with a non-zero component of this ancestry. We show groups modelled in c that have a feasible model (P > 0.01). In c,d, we show one s.e. BA, Bronze Age; CNE, continental northern Europeans; EBA, early Bronze Age; EVA, early Viking Age; IA, Iron Age; MED, medieval; MLBA, middle/late Bronze Age; VA, Viking Age.