Fig. 3

Detection of fetal trisomy T21 using CG-fraction. a Effect of fetal fraction on the Z-score indicating fetal karyotype. Normalized CG-fractions in chromosome 21 were used to compute Z-scores. 100% diagnostic accuracy of fetal trisomy T21 independent of fetal fraction was achieved. Dashed lines represent logistic regression fits. hmTOP-seq samples of shallow sequencing (2.5 million raw reads) are indicated with red circles. b Effect of reduced library size on classification accuracy. 100% diagnostic accuracy is achieved using 3 million or 1 million of processed sequencing reads of uTOP-seq and hmTOP-seq, respectively. Reads were randomly sampled and karyotype determined using leave-one-out cross-validation. In each cross-validation loop a logistic regression model was built with the Z-scores computed from the normalized CG-fraction within chromosome 21. Error bars indicate the standard deviation from mean AUC across 30 sampling iterations