The Fellow sampled from an existing collection of archaeological samples at the University College Dublin, Ireland and sampled on site in western Ukraine. Samples, human petrous bone, were used to infer differences in endogenous DNA yield from warm and humid or warm and arid environments, which are particularly challenging for researchers in the field of human ancient DNA. The reasons for the lack of quality DNA from these environments are varied, ranging from soil pH to burial context. We employed a sub-sampling approach to the inner osseous canal of the petrous bone, as this has consistently shown to yield higher amounts of human DNA compared to other skeletal elements. However, researchers have only been able to obtain small (less than 1%) of endogenous DNA from these samples; or use expensive technologies such as in-solution or array-based target capture to obtain higher yields from these difficult samples. We developed a protocol that targets several regions within the inner osseous canal that we hope will benefit future researchers who work with such samples. Preliminary results show that the most successful region for obtaining quality DNA is the cochlea; however, intra-sample and intra-site variability make observing consistent patterns difficult. Therefore, we suggest a case-by-case basis for archaeological samples deriving from these environments. Our sampling approach has yielded high endogenous DNA that will bring down the costs for future researchers. The Fellow undertook several experiments in parallel with his sub-sampling strategy that implemented how NGS library techniques might influence sequencing complexity and endogenous DNA yield. To this end, he constructed over 500 NGS libraries, of which over 250 were sequenced on Illumina platforms. Specifically, he prepared libraries according to a traditional approach, a modified traditional approach, and a technique called the single-stranded method that has successfully been used to sequence Neanderthal and Denisovan samples. The single-stranded method has also been shown to work well for highly degraded samples of which the Fellow was working with. To this end, he built several libraries according to these different methods from the same extract and found that only the most degraded samples benefit from the single-stranded approach. This is important, as the single-stranded approach is costlier than the more traditional double-stranded methods and more time consuming. Therefore, for samples that are relatively of good quality, the less expensive and higher throughput double-stranded approach is best.
