A method for positive forensic identification of samples from extremely low-coverage sequence data

Background: Determining whether two DNA samples originate from the same individual is difficult when the amount of retrievable DNA is limited. This is often the case for ancient, historic, and forensic samples. The most widely used approaches rely on amplification of a defined panel of multi-allelic markers and comparison to similar data from other samples. When the amount retrievable DNA is low these approaches fail. Results: We describe a new method for assessing whether shotgun DNA sequence data from two samples are consistent with originating from the same or different individuals. Our approach makes use of the large catalogs of single nucleotide polymorphism (SNP) markers to maximize the chances of observing potentially discriminating alleles. We further reduce the amount of data required by taking advantage of patterns of linkage disequilibrium modeled by a reference panel of haplotypes to indirectly compare observations at pairs of linked SNPs. Using both coalescent simulations and real sequencing data from modern and ancient sources, we show that this approach is robust with respect to the reference panel and has power to detect positive identity from DNA libraries with less than 1 % random and non-overlapping genome coverage in each sample. Conclusion: We present a powerful new approach that can determine whether DNA from two samples originated from the same individual even when only minute quantities of DNA are recoverable from each.