Fig 1.
Trisomy detection accuracy of tested NIPT software tools across different sequencing depths.
(A) Percentages and absolute numbers of undetected trisomy cases among known trisomies. (B) False-positive T21 results among known euploid samples. The horizontal dashed line in each graph marks the 1% cut-off level often used in clinical screening.
Fig 2.
Z-scores of clinically validated T21 samples across a range of sequencing depths.
Z-scores of known trisomy samples at sequencing depths of 20M RPS (A), 15M RPS (B), 10M RPS (C), 5M RPS (D), 2.5M RPS (E), and 1.25M RPS (F) are shown. Undetected (false negative) trisomies falling below the Zt cut-off thresholds (black dashed line in each graph) are represented as black triangles.
Fig 3.
Correlation between the ‘true’ (based on 20M RPS data) and low sequencing depth based FF estimates.
Pearson correlation data are shown for 20M RPS FF 0–5% estimates and estimates obtained at sequencing depths of 15M RPS (A), 10M RPS (B), 5M RPS (C), 2.5M RPS (D), and 1.25M RPS (E).
Fig 4.
The effect of FF on detection of T21 across different sequencing coverages.
Ze cut-off was used for identifying the presence of the trisomy (internal classification in the case of GIPseq). Black triangles represent undetected trisomy cases. The 20M RPS group served as the standard for FF calculations. Data obtained with sequencing depths of 20M RPS (A), 15M RPS (B), 10M RPS (C), 5M RPS (D), 2.5M RPS (E), and 1.25M RPS (F) are shown.
Table 1.
The key considerations for each of the compared algorithms.
Table covers minimal coverage where the computational tool could simultaneously detect trisomy cases in chromosomes 13, 18 and 21 with less than 1% undetected and false-positive trisomy calls, minimal fetal fraction from which upwards there are no undetected T13, T18 and T21 calls, and software tool usability.