| dc.description.abstract | Sequencing DNA derived from ancient bones (‘aDNA’) enabled direct genetic comparison of anatomically modern humans (AMHs) and their ancestral populations, as well as archaic hominins like Neanderthals and Denisovans. From this, we know that archaic hominins interbred with ancient AMHs, and have a more detailed understanding of population movements throughout history. However, interpreting what genetic differences between groups imply about phenotypic differences remains challenging. To do this, we adapted an approach for imputing gene regulatory differences based on individuals’ genotypes for application in the context of ancient DNA.
First, to potential gene regulatory implications of Neanderthal introgression into ancient AMHs, we applied this approach to high-quality genomes from archaic hominins. We found 766 genes that are likely to have been divergently regulated (DR) by Neanderthal haplotypes that do not remain in AMH populations. DR genes include many involved in skeletal phenotypes known to differ between Neanderthals and AMHs, and are also enriched for genes associated with spontaneous abortion, polycystic ovary syndrome, myocardial infarction and melanoma. Comparing DR genes between two Neanderthals and a Denisovan revealed divergence in the immune system and in genes associated with skeletal and dental morphology that are consistent with the archaeological record.
However, to apply a similar approach to more recent evolutionary questions about AMH populations, we needed to consider the fact that most aDNA genomic information is low-coverage, which would impact the ability of our approach to make accurate gene regulatory predictions. We therefore used simulated genomes and populations to better understand how gene regulatory models behave with aDNA-derived data, and to develop new approaches to increase their robustness to incomplete data. We found that, while prediction accuracy decreased as the missingness in genomes increased, this effect could be minimized by training models specifically using variants that were reliably captured in the genome data.
Using the knowledge gained from the simulations, we then applied targeted gene regulatory models to 490 ancient Eurasians to study differences in populations with differing lifestyles. We identified thousands of genes with significant regulatory differences among the three groups. Our results recapitulated known instances of altered gene expression relevant to population changes diet and lifestyles, such as FADS1, and they also suggested explanations for previously observed signals of selection on LEPR. In general, genes involved in metabolic and immune pathways were the most enriched among these genes, indicating that these pathways are the most likely to have been affected by altered gene regulation during recent human evolution.
This work provides an avenue for exploring phenotypic differences between ancient groups from genomic information alone, and demonstrates its use in connecting genomic differences to phenotypic differences. | |
