Extended Data Fig. 1: Global ancestry and homozygosity.

A. Global ancestry proportions for NCIG, Papuan and 1000 Genomes. The software ADMIXTURE run with 159 Indigenous Australian samples, 60 Papuan and 2,600 samples from the LC 1000 Genomes. Samples shown horizontally by population (with reduced bar width for the 1000 Genomes samples) and cluster (colour) assignment proportion shown as bar height. ADMIXTURE was run assuming the sample contained from k = 2 up to k = 12 clusters (y-axis). No ancestry mask applied. Restricted to biallelic SNVs, MAF > 0.01 and LD thinned. This analysis was used to estimate non-Indigenous Australian or PNG ancestry in the NCIG and PNG samples. B. Runs of homozygosity (ROH) for the NCIG + PNG (unmasked) dataset and a subset of (HC) 1000 Genomes samples. Mean count versus mean sum of ROH segments greater than 1 Mb in length. Error bars are within population SEM. Note that a long-term reduction in effective population size is expected to increase both the count and total length of ROH (as seen for NCIG populations), whereas recent consanguinity generates a small number of long ROH. C. Per-individual ROH length (for ROH > 1 Mb) as a percentage of the total autosomal genome length (2.8 Gb). Individuals from three Indigenous populations from Nepal (Kusunda) and Brazil (Surui and Karitiana) from the SGDP were included for comparison as some exhibited extreme levels of ROH. Individuals identified as “Tiwi-outliers” were included in the Tiwi sample in this plot and are evident in panel C as a cluster of individuals with reduced ROH.