Skip to main content
. 2021 Apr 19;13:82. doi: 10.1186/s13148-021-01055-z

Fig. 1.

Fig. 1

Analysis flow. START newborns were grouped into ART and non-ART, and each group was randomly assigned to a training and test set. The non-ART training set was used to develop the EPIC GA clock and the 450 K/EPIC overlap clock. The ART training set was used to develop the ETD-based clock. All three clocks were tested in the non-ART test set. The EPIC GA clock, the Bohlin clock, and the Knight clock were also tested in the PREDO test set. The datasets are marked in green, and the clocks are marked in blue. START-derived datasets and clocks are marked with solid lines. External datasets and clocks are marked with dashed lines