Adaptable Model Parameters in Non-Invasive Prenatal Testing Lead to More Stable Predictions

Area of Science:

• Genetics
• Molecular Biology
• Reproductive Health

Background:

• Massively parallel shotgun sequencing enables non-invasive prenatal testing (NIPT) for fetal aneuploidy using maternal plasma DNA.
• Current NIPT methods compare chromosomal distributions against control sets, with accuracy dependent on sequenced fragment counts.
• Variability in sequencing depth between tested samples and control sets can lead to inaccurate variance estimation and false positive/negative results.

Purpose of the Study:

• To develop a more robust method for NIPT by addressing the limitations of fixed control set variances.
• To improve the reliability and accuracy of non-invasive prenatal testing for fetal aneuploidy.
• To enhance diagnostic stability in NIPT, particularly in real-world scenarios with variable sequencing coverage.

Main Methods:

• Proposed an adaptive calculation of variance for each tested NIPT sample.
• Based the adaptive variance calculation on the specific number of sequenced fragments within each individual sample.
• Evaluated the method's performance in diagnostics with highly divergent inter-sample coverage.

Main Results:

• The adaptive variance calculation demonstrated more stable predictions compared to traditional methods.
• The proposed approach mitigates over- or under-estimation of variance caused by differing sequencing depths.
• Improved diagnostic accuracy in real-world NIPT applications with variable sequencing data.

Conclusions:

• Adaptive variance calculation is a key improvement for non-invasive prenatal testing accuracy.
• This method enhances the reliability of fetal aneuploidy detection from maternal plasma DNA.
• The findings support the adoption of adaptive variance for more precise NIPT diagnostics.