Noninvasive prenatal diagnosis of fetal trisomy 18 and trisomy 13 by maternal plasma DNA sequencing

Abstract

Massively parallel sequencing of DNA molecules in the plasma of pregnant women has been shown to allow accurate and noninvasive prenatal detection of fetal trisomy 21. However, whether the sequencing approach is as accurate for the noninvasive prenatal diagnosis of trisomy 13 and 18 is unclear due to the lack of data from a large sample set. We studied 392 pregnancies, among which 25 involved a trisomy 13 fetus and 37 involved a trisomy 18 fetus, by massively parallel sequencing. By using our previously reported standard z-score approach, we demonstrated that this approach could identify 36.0% and 73.0% of trisomy 13 and 18 at specificities of 92.4% and 97.2%, respectively. We aimed to improve the detection of trisomy 13 and 18 by using a non-repeat-masked reference human genome instead of a repeat-masked one to increase the number of aligned sequence reads for each sample. We then applied a bioinformatics approach to correct GC content bias in the sequencing data. With these measures, we detected all (25 out of 25) trisomy 13 fetuses at a specificity of 98.9% (261 out of 264 non-trisomy 13 cases), and 91.9% (34 out of 37) of the trisomy 18 fetuses at 98.0% specificity (247 out of 252 non-trisomy 18 cases). These data indicate that with appropriate bioinformatics analysis, noninvasive prenatal diagnosis of trisomy 13 and trisomy 18 by maternal plasma DNA sequencing is achievable.

Figure 1. Recruitment of samples.

Figure 2. Z-scores of chromosome 13 and 18 determined by the standard z-score approach and the z-score approach with GC correction.

Z-scores of (A) chromosome 13 and (B) chromosome 18 determined by the standard z-score approach with alignment against the repeat-masked or non-repeat-masked reference genomes and the z-score approach with GC correction are shown. Horizontal dashed lines indicate the z-score cut off value of 3. T13, trisomy 13. T18, trisomy 18.

Figure 3. Precision before and after GC correction for the autosomes.

Coefficient of variation (CV) for each chromosome was calculated based on the control euploid cases before GC correction with alignment against the repeat-masked or non-repeat-masked reference genomes and after GC correction with alignment against the non-repeat-masked reference genome. Chromosomes are ordered from left to right in increasing GC contents. GC content of each chromosome is shown in the brackets. 95% confidence interval was shown in error bars.