(A) Map of Europe showing the Iberian sampling sites (red circles, size proportional to sample sizes) and the sites for published aurochs genomes used in the analysis (blue diamonds). (B) NMDS ordination of nuclear data of Iberian samples considered B. taurus samples (orange), Iberian samples that were morphologically considered B. primigenius (red), other Iberian Bos samples (brown), ancient domestic cattle from the Balkans and Anatolia (yellow), and aurochs (blue). Data point shape corresponds to mitochondrial haplogroups. (C) f4 statistic measuring allele sharing of the Iberian samples with European aurochs (CPC98) or Anatolian Neolithic cattle (Sub1). The time periods displayed are contextual.

Ancestry proportions in all sequenced Iberian Bos samples. For each sample, only feasible and not rejected (p>0.01) models are displayed. For moo034 and moo010p, both single source models are consistent with the data, we display the single source model with the higher p-value. Error bars are showing block-jackknife estimates of the 95% confidence interval.

(A) Estimates of aurochs ancestry (estimated using the 2 source model, if working and except for Mesolithic individuals) in directly dated Iberian samples over time. Error bars indicate one standard error for the 2 source models. Modern Iberian breeds with working 2 source models are added around date 0 with some vertical jitter. Approximate boundaries for the main sampling periods are indicated by dashed vertical lines. (B) Comparison of allele sharing of Iberian samples with aurochs and Anatolian domestic on the autosomes versus X chromosomes. Individuals were separated into three separate groups: Mesolithic samples, samples with admixed ancestry (sorted by autosomal f4 estimate), and individuals for which only the single source, domestic model fit the data. (C) qpAdm estimates of Aurochs ancestry in modern western European cattle breeds from the (26) dataset. Commercial breeds are marked with a “C”. The figure is only showing breeds with feasible and non-rejected two source models, all results are shown in Supplementary Information, Table S2. Error bars are showing block-jackknife estimates of the 95% confidence interval.