Adaptive parameter of standard deviation enhances the power of noninvasive prenatal screens

Traditional Z-test methods during noninvasive prenatal screens (NIPS) use the fixed parameter of standard deviation (SD), which ignores the influence of actual sequencing read counts of a sample on the results. The aim of this study is to eliminate the influence of the sequencing depth of individual samples on the results and enhance the power of NIPS.
In this study, we propose an improved NIPS method, which calculates the SD in the Z-score process adaptively according to the actual read count of the test sample. Our approach obtained the SD linear fitting function along with the read count with a large number of reference samples, in which SD and read count fit well. The effectiveness of our enhanced NIPS method was evaluated on three common trisomy syndromes and five recurrent CNV syndromes with 3219 and 6592 samples based on whole genome sequencing of maternal peripheral blood.
A total of 3,219 pregnant samples have been used for validating the proposed method on detecting fetal trisomy syndromes (T13, T18, and T21), in which eight false negative (FN) samples have been corrected as true positive (TP) and eight false positive (FP) samples have been fixed as true negative (TN) with our proposed adaptive-SD method. Another 6592 samples were used to compare the two methods on detecting five recurrent fetal copy number variation (CNV) syndromes, in which the FP samples have decreased from 99 to 39.
Our adaptive-SD NIPS method shows more power on detecting both trisomy syndromes and five recurrent CNVs in the pregnant samples with diverse read counts. Besides, our proposed method contributes to lower FP and FN samples than the traditional Z-test method in NIPS. Our results show that our enhanced NIPS methods are effective in detecting both abnormal fetal trisomy syndromes and recurrent CNV syndromes in pregnant women.

无创产前筛查（noninvasive prenatal screens, NIPS）中传统Z检验方法采用固定标准差（standard deviation, SD）参数，忽略了样本实际测序读段计数对结果的影响。本研究旨在消除单一样本测序深度对结果的干扰，提升无创产前筛查的检测效能。
本研究提出一种改进的无创产前筛查方法，该方法可根据待测样本的实际读段计数，在Z得分计算过程中自适应计算标准差。本方法通过大量参考样本建立标准差与读段计数的线性拟合函数，二者拟合度优异。本研究基于孕妇外周血全基因组测序数据，针对三种常见三体综合征及五种复发性拷贝数变异（copy number variation, CNV）综合征，分别纳入3219例和6592例样本，对改进后的无创产前筛查方法的有效性进行评估。
共纳入3219例妊娠样本，用于验证所提方法在胎儿三体综合征（T13、T18、T21）检测中的性能：采用本自适应标准差方法后，8例假阴性（false negative, FN）样本被修正为真阳性（true positive, TP），8例假阳性（false positive, FP）样本被修正为真阴性（true negative, TN）。另外纳入6592例样本，用于对比两种方法在五种复发性胎儿拷贝数变异综合征检测中的效果：经本方法处理后，假阳性样本数量从99例降至39例。
本自适应标准差无创产前筛查方法，在不同读段计数水平的妊娠样本中，对三体综合征及五种复发性拷贝数变异均展现出更优的检测效能。相较于无创产前筛查中传统Z检验方法，本方法可有效降低假阳性与假阴性样本的数量。研究结果证实，改进后的无创产前筛查方法在检测孕妇胎儿异常三体综合征及复发性拷贝数变异综合征方面均具有良好有效性。