Optimized in-solution enrichment of over a million ancient human SNPs

Abstract

Background: In-solution hybridization enrichment of genetic markers is a method of choice in paleogenomic studies, where the DNA of interest is generally heavily fragmented and contaminated with environmental DNA, and where the retrieval of genetic data comparable between individuals is challenging. Here, we benchmark the commercial "Twist Ancient DNA" reagent from Twist Biosciences using sequencing libraries from ancient human samples of diverse demographic origin with low to high endogenous DNA content (0.1-44%). For each library, we tested one and two rounds of enrichment and assessed performance compared to deep shotgun sequencing.

Results: We find that the "Twist Ancient DNA" assay provides robust enrichment of approximately 1.2M target SNPs without introducing allelic bias that may interfere with downstream population genetics analyses. Additionally, we show that pooling up to 4 sequencing libraries and performing two rounds of enrichment is both reliable and cost-effective for libraries with less than 27% endogenous DNA content. Above 38% endogenous content, a maximum of one round of enrichment is recommended for cost-effectiveness and to preserve library complexity.

Conclusions: In conclusion, we provide researchers in the field of human paleogenomics with a comprehensive understanding of the strengths and limitations of different sequencing and enrichment strategies, and our results offer practical guidance for optimizing experimental protocols.

Keywords: Ancient DNA; Enrichment; Human paleogenomics; Population genetics.

Fig. 1

Twenty-four aDNA libraries, representing samples originating from Iberia, Southeast Asia, and Central America, were selected, all underwent deep shotgun sequencing, and were also input into enrichment experiments. This included six unpooled libraries that were enriched individually, while the remaining libraries were pooled into six distinct reactions: 3 reactions with 2 low endo% DNA libraries each and 3 reactions with 4 high endo% DNA libraries each. All sets of pooled and unpooled libraries underwent one as well as two rounds of enrichment prior to sequencing

Fig. 2

Enrichment efficacy of deep shotgun sequencing (SG) or Twist enrichment (either one or two rounds, i.e., TW1 or TW2, respectively) in relation to the mappable endo% from the screening shotgun data (x-axis). Enrichment efficacy was measured using A total on-target SNPs per sample, B SNPs per million sequenced read pairs, C SNPs per million reads going into mapping, D SNPs per million mapped reads, E SNPs per million filtered reads, passing filter for MAQ > 25, F SNPs per million unique reads, G sequenced post-enrichment endo%, H mappable post-enrichment endo%, I filtered post-enrichment endo%, and J unique post-enrichment endo%. Point shape and color corresponds to the method. Gray vertical lines connect points showing the three methods performed on the same library. The solid lines and shaded areas show fitted linear regression models (shotgun) or logarithmic transformed linear regression models (Twist enrichments) as appropriate. The black line in G–J shows y = x

Fig. 3

A Cost per SNP (AUD) obtained from shotgun sequencing and Twist enrichment methods as a function of the mappable endo%. The y-axis is log10 transformed. B Relative fold cost saving per SNP (AUD) obtained from TW1 and TW2 compared to deep shotgun sequencing. The r² values represent the correlation value of the data points with the fitted model trendline. Point shape and color corresponds to the method. Gray vertical lines connect points showing the methods performed on the same library. Linear or log10 transformed linear models are fit to each method (solid lines), with 95% confidence intervals (shaded areas). This figure is also available as an interactive app where users can input different costs of the Twist enrichment kit or sequencing to predict how the savings will change with different prices. See https://roberta-davidson.github.io/Davidson_etal_2024-Twist/ for the interactive plot

Fig. 4

Cost per SNP in relation to SNP coverage (x-axis) and mappable endo% measured from screening shotgun data (colors) for deep shotgun sequencing and enrichment methods (shapes). A Total SNP coverage on Twist ancient DNA panel and B SNPs per million sequenced read pairs. Lines connect the same library across the three methods. Plot axes and color scale are log10 transformed

Fig. 5

f₄ statistics of the form f₄ (Mbuti, Pop1.MethodA; Pop2.MethodA, Pop2.MethodB). Pop 1 and 2 refer to the geographic population used in each statistical calculation and are faceted horizontally and vertically, respectively. Method A and B refer to the comparison of two of the three methods (SG = shotgun, TW1 = 1 round of Twist enrichment, TW2 = 2 rounds of Twist enrichment); the methods analyzed are annotated on the left y-axis. Thick and thin error bars represent 2 and 3 standard error deviations (s.e.), respectively. Tests where |Z|> 2 are colored red, no test returned |Z|> 3. Tests with < 10,000 SNPs are annotated with the number of SNPs

Fig. 6

Equity of library enrichment in pooled reactions. Library pools are ordered from low to high endo% from top to bottom. Color denotes one library and each library is labeled in the first column with the mappable endo% calculated from shotgun screening sequencing. Panels show from left to right: number of sequenced read pairs, number of reads input to mapping, number of reads mapped, number of mapped reads passing mapping quality (≥ q25), and deduplicated mapped reads

Fig. 7

Pairwise rate of single index hopping between every pair of libraries, categorized into (left to right) 2-library pools, 4-library pools, and all other pairs of libraries. Each category is annotated with the number of pairs (n). The hopping rate was calculated as the proportion of reads with a hopped index combination over the sum of reads from every possible index combination within the pair

Fig. 8

Detailed workflow of the experimental design for Twist enrichment laboratory protocol used in this study, showing the steps conducted on day 1 and day 2, with the hybridization incubation overnight between the days