Abstract
Advances in technology have decreased the cost of generating genetic data from ancient people, resulting in exponentially increasing numbers of individuals with whole-genome data. However, these technologies come with platform-specific biases, limiting coanalyzability of individuals sequenced with different technologies as well as joint analysis of modern and ancient individuals. Here, we present a method to identify single-nucleotide polymorphisms (SNPs) with minimal technology-specific bias. Leveraging data from more than 18,000 individuals, we apply this method to identify a set of around 1 million SNPs that we call the “compatibility” panel, which has been effectively assayed in a large fraction of ancient human DNA experiments published to date. We also identify a subset of these SNPs, the “compatibility-HO” panel, which are restricted to positions that have been assayed in more than 10,000 modern people from more than 1000 diverse populations using the Affymetrix Human Origins (HO) genotyping array. The compatibility panel reduces spurious Z-scores owing to differing sequencing platforms by nearly an order of magnitude, while retaining ∼60%–85% of statistical power for f-statistic analysis. We also provide a tool for users to select different tradeoffs between bias and power as well as sequencing platforms for their specific analyses.