AKmerBroom: Ancient oral DNA decontamination using Bloom filters on k-mer sets

Abstract

Dental calculus samples are modeled as a mixture of DNA coming from dental plaque and contaminants. Current computational decontamination methods such as Recentrifuge and DeconSeq require either a reference database or sequenced negative controls, and therefore have limited use cases. We present a reference-free decontamination tool tailored for the removal of contaminant DNA of ancient oral sample called aKmerBroom. Our tool builds a Bloom filter of known ancient and modern oral k-mers, then scans an input set of ancient metagenomic reads using multiple passes to iteratively retain reads likely to be of oral origin. On synthetic data, aKmerBroom achieves over sensitivity and specificity. On real datasets, aKmerBroom shows higher read retainment ( on average) than other methods. We anticipate aKmerBroom will be a valuable tool for the processing of ancient oral samples as it will prevent contaminated datasets from being completely discarded in downstream analyses.