Ancient DNA: from methodological challenges to natural selection and ancient mobility

The study of ancient DNA (aDNA) allows to recover information on past human populations and how they evolved, moved, and mixed across time. The present thesis contains four chapters covering different subjects in the aDNA field. In chapter 1, we investigate the batch effect that arises when mixing different types of data. Most ancient human genomes are generated either by whole genome shotgun sequencing, or by target enrichment of 1240K SNPs. We show that when co-analysing genomes of both types in large datasets, an important bias affects state of the art methods such as MDS (starting dimension 3), ADMIXTURE (starting K=4), and D-statistics. We show that removing SNPs statistically associated to data type with $p