Allelic bias when performing in‐solution enrichment of ancient human DNA

Abstract

In‐solution hybridisation enrichment of genetic variation is a valuable methodology in human paleogenomics. It allows enrichment of endogenous DNA by targeting genetic markers that are comparable between sequencing libraries. Many studies have used the 1240k reagent—which enriches 1,237,207 genome‐wide SNPs—since 2015, though access was restricted. In 2021, Twist Biosciences and Daicel Arbor Biosciences independently released commercial kits that enabled all researchers to perform enrichments for the same 1240 k SNPs. We used the Daicel Arbor Biosciences Prime Plus kit to enrich 132 ancient samples from three continents. We identified a systematic assay bias that increases genetic similarity between enriched samples and that cannot be explained by batch effects. We present the impact of the bias on population genetics inferences (e.g. Principal Components Analysis, ƒ‐statistics) and genetic relatedness (READ). We compare the Prime Plus bias to that previously reported of the legacy 1240k enrichment assay. In ƒ‐statistics, we find that all Prime‐Plus‐generated data exhibit artefactual excess shared drift, such that within‐continent relationships cannot be correctly determined. The bias is more subtle in READ, though interpretation of the results can still be misleading in specific contexts. We expect the bias may affect analyses we have not yet tested. Our observations support previously reported concerns for the integration of different data types in paleogenomics. We also caution that technological solutions to generate 1240k data necessitate a thorough validation process before their adoption in the paleogenomic community.

Tales from the end of the world: three decades of paleogenetic research in Patagonia

“…However, while it is not clear this benefits research groups equally [135], there has also been increasing concerns about possible biases in the data compared to shotgun sequencing, including a strong reference bias [136]. Additionally, a systematic assay bias that induces higher genetic similarity between individuals enriched with the same kit has been suggested [137]. These biases may lead to erroneous conclusions when the target populations are highly genetically differentiated from the populations used in the ascertainment scheme to build the capture array, which is the case with Patagonian populations.…”

GRUPS-rs, a high-performance ancient DNA genetic relatedness estimation software relying on pedigree simulations

“…It must be pointed out that the benchmark and case-study results presented here make extensive use of the EUR population and the 1240K SNP panel, a dataset which is still widely used during paleogenomic analyses to this day, but is now known to be a potential source of ascertainment bias, particularly when applied against ancient non-Eurasian populations [57,58,59]. Hence, we caution that the raw accuracy values showcased in section 3.3 are expected to vary according to the source population at hand, and the targeted panel of SNPs, as previously noted in [9].…”