snpAD: an ancient DNA genotype caller

Abstract

Motivation: The study of ancient genomes can elucidate the evolutionary past. However, analyses are complicated by base-modifications in ancient DNA molecules that result in errors in DNA sequences. These errors are particularly common near the ends of sequences and pose a challenge for genotype calling.

Results: I describe an iterative method that estimates genotype frequencies and errors along sequences to allow for accurate genotype calling from ancient sequences. The implementation of this method, called snpAD, performs well on high-coverage ancient data, as shown by simulations and by subsampling the data of a high-coverage Neandertal genome. Although estimates for low-coverage genomes are less accurate, I am able to derive approximate estimates of heterozygosity from several low-coverage Neandertals. These estimates show that low heterozygosity, compared to modern humans, was common among Neandertals.

Availability and implementation: The C++ code of snpAD is freely available at http://bioinf.eva.mpg.de/snpAD/.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Schematic overview of the method implemented in snpAD

Fig. 2.

Accuracy of parameter estimation for simulated datasets. Left: Deviation from simulated genotype probabilities for the six heterozygous genotypes. Each simulation is indicated by a vertical blue dotted line and the estimates are shown as blue points. Estimates for 3-fold coverage deviated by more than 0.5 and are not visible at the depicted range. Right: Average deviation from simulated error probabilities

Fig. 3.

Parameter estimates for subsampled Vindija 33.19 data compared to full data. Left: Estimated genotype frequencies (points) compared to full data (shown as horizontal lines). Right: Average deviation from error rates in the full dataset. Estimates for 1-fold coverage fall outside of the plotted ranges

Fig. 4.

Genotype frequencies for autosomal sites with at least 4-fold coverage. Left: Vindija 33.19 full data and data from a single subsampled library. Violin plots show the distribution over Vindija 33.19 chromosomes. Right: Low-coverage Neandertals. Horizontal lines show genome-wide Vindija 33.19 estimate