S6 Table -

Performance of ascertainment schemes on simulated data (those shown in Fig 3) explored across three population quintuplets (including either “Denisovan”, or “Neanderthal 1”, or “Neanderthal 2” “archaic” individuals, in addition to the “African 1”, “African 2”, “non-African 1”, and “non-African 2” groups) and assessed as the fraction of all topologies that are rejected under ascertainment (fit poorly with WR >3 SE) but accepted on all sites (fit well with WR <3 SE), or as the fraction of all topologies that are accepted under ascertainment (WR <3 SE) but rejected on all sites (WR >3 SE). We also applied the binary classifier (based on a 10th percentile threshold and 10 random site subsamples matching the average size of the HO one-panel set, 500K sites) to determine if the ascertainment produces unbiased or biased results (the latter cases are highlighted in bold and underlined text). Ten independent simulations with the same parameters were performed, and the following ascertainment schemes were explored on each of them: 1) archaic ascertainment (1.05M sites on average across simulation iterations); 2) HO one-panel ascertainment, based on the “African 2” group (500K sites on average); 3) HO four-panel ascertainment (based on the “African 1”, “African 2”, “non-African 1”, and “non-African 2” groups, 1.34M sites on average); 4) AFR MAF ascertainment, that is restricting to sites with MAF >5% in the union of the “African 1” and “African 2” groups (1.85M sites on average); 5) global MAF ascertainment on the union of the “African 1”, “African 2”, “non-African 1”, and “non-African 2” groups (1.62M sites on average, abbreviated as “AMH MAF”); 6) non-African MAF ascertainment (1.48M sites on average).
(XLSX)