Ancient comparative datasets

The 545,942 SNPs retained in the rWD1560-dataset (see “Reduced worldwide (rWD1560) dataset” section) were called on the ancient dataset that contained our 12 ancient Panamanian individuals merged with 241 ancient Siberian and American individuals with a minimum coverage of 0.01X (Table S3). The calling was performed for all individuals in one run using ANGSD (Korneliussen et al., 2014) with the haplocall 1 option, which picks a random read starting from an input set. In addition, to avoid possible biases due to low coverage data, we down-sampled all ancient genomes to 1X and 0.5X coverage using ANGSD with the -downSample option.

The ancient dataset was merged using PLINK 1.9 (Purcell et al., 2007) with our modern datasets and then filtered using–geno and–mind options set respectively to 0.60 and 0.98 (Scheib et al., 2018), keeping only individuals with at least 10,000 SNPs (Table S3) (Posth et al., 2018).