A catalog of ancient proxies for modern genetic variants

The ability to observe the genomes of past human populations using ancient DNA provides an extraordinary perspective on many fundamental questions in human genetics, including understanding the evolutionary history of variants that underlie human disease and other phenotypes. However, ancient DNA is often damaged and degraded, yielding low-coverage of most nucleotides. Further, many publicly available genotypes for ancient humans are limited to ~1.23 million specific loci. Thus, variants of interest often fall outside these specific positions, limiting the ability of ancient DNA to shed light on many loci. Here, we address this challenge by quantifying linkage disequilibrium (LD) between modern variants and ancient genotyped variants (AGVs) to generate a catalog enabling rapid identification of proxy variants. We identified 260,732,675 pairs of AGVs and modern variants with a minimum LD threshold hold of R ² ≥ 0.2. Even at R ² ≥ 0.9, ≥ 60% of common variants were linked to an AGV in non-African ancestry groups, as were 34% of common variants in Africans. We evaluated the accuracy of the genotypes inferred from proxy variants in two high-coverage ancient genomes finding that > 90% of genotypes were correctly predicted, even in a 45,000 year old individual. We also find that AGVs are significantly older than expected and that many likely are evolving neutrally. We integrate these results in a database that researchers can easily query to identify ancient proxy variants if their variant of interest is not directly genotyped in ancient humans.