Evaluation of genotype imputation on ancient Anatolian human genomes

Genotype imputation presents a solution to the high amount of missing information in low coverage ancient DNA samples. In this thesis, by using the phasing and imputation tool GLIMPSE, the accuracy of genotype imputation and the amount of missing variants on autosomal biallelic SNPs are evaluated using different genotype posterior probability filters at four different coverages using ten high coverage and over a hundred low coverage genomes from Anatolia and West Eurasia. The effect of using imputed data in frequently used population genetics analyses is investigated. The accuracy of imputation measured as genotype concordance with high coverage data reaches 70% to 90% in common variants at 0.1x genome coverage while losing less than 5% to 15% of all variants using filters between 80% and 99%. Missing amount in common variants using the same data reaches %20 and %50 in commonly used Human Origins dataset positions. Pseudohaploid data have more affinity towards the reference genome compared to the imputed data, pointing to mitigation of reference bias by imputation. On the other hand, imputed data show tendency towards the reference panel populations as the coverage decreases. Allele frequency calculations using imputation and alignment data show up to 15% difference in low coverage Anatolian Neolithic samples. Results show that it is possible to get reliable results using imputed data for common variants for low coverage ancient genome samples from the region of interest.