Fig. 6

Classification of aDNA and present-day DNA sequences. a The receiver operating characteristic (ROC) curves illustrate the performance of AuthentiCT (solid) and PMDtools (dashed) to identify aDNA sequences. A sequence is considered ancient if the log-likelihood ratio (score) of an ancient versus present-day origin is equal to or higher than a threshold (different colours). Each point represents the average performance over 19 datasets (of 10,000 sequences each) with varying proportions of ancient and present-day DNA sequences (5 to 95% in steps of 5%) for AuthentiCT (circles), PMDtools (squares) and a filter for sequences exhibiting at least one C-to-T substitution within the first or last three positions (deam. filter; triangles). Sequences are from Mezmaiskaya 2 (libraries A9180, A9288, A9289 and R1917 [58]) and the present-day human control. The bars correspond to two standard errors. b–d The distributions illustrate the number of C-to-T substitutions per sequences (b), the distance between a C-to-T substitution and the closest end of the sequence (c) or the closest non-deaminated base (d) in sequences classified as ancient only by AuthentiCT (left), only by PMDtools (middle) or both (right), using a score threshold of 3 as recommended for PMDtools [41]