Li Jia, Ju Jia, Zhao Qiang, Liu Weiqiang, Yuan Yuying, Liu Qiang, Zhou Lijun, Han Yuan, Yuan Wen, Huang Yonghua, Xie Yingjun, Li Zhihua, Chen Jingsi, Huang Shuyu, Chen Rufang, Li Wei, Tan Meihua, Wang Danchen, Zhou Si, Zhang Jian, Zeng Fanwei, Yu Nan, Su Fengxia, Chen Min, Ge Yunsheng, Huang Yanming, Jin Xin
BGI Genomics, BGI-Shenzhen, Shenzhen, China.
Hebei Industrial Technology Research Institute of Genomics in Maternal & Child Health, Shijiazhuang BGI Genomics Co., Ltd., Shijiazhuang, China.
Front Genet. 2022 Feb 10;13:802865. doi: 10.3389/fgene.2022.802865. eCollection 2022.
The existence of maternal malignancy may cause false-positive results or failed tests of NIPT. Though recent studies have shown multiple chromosomal aneuploidies (MCA) are associated with malignancy, there is still no effective solution to identify maternal cancer patients from pregnant women with MCA results using NIPT. We aimed to develop a new method to effectively detect maternal cancer in pregnant women with MCA results using NIPT and a random forest classifier to identify the tissue origin of common maternal cancer types. For examination, 496 participants with MCA results via NIPT were enrolled from January 2016 to June 2019 at BGI. Cancer and non-cancer participants were confirmed through the clinical follow-up. The cohort comprising 42 maternal cancer cases and 294 non-cancer cases enrolled from January 2016 to December 2017 was utilized to develop a method named mean of the top five chromosome z scores (MTOP5Zscores). The remaining 160 participants enrolled from January 2018 to June 2019 were used to validate the performance of MTOP5Zscores. We established a random forest model to classify three common cancer types using normalized Pearson correlation coefficient (NPCC) values, z scores of 22 chromosomes, and seven plasma tumor markers (PTMs) as predictor variables. 62 maternal cancer cases were confirmed with breast cancer, liver cancer, and lymphoma, the most common cancer types. MTOP5Zscores showed a sensitivity of 85% (95% confidence interval (CI), 62.11-96.79%) and specificity of 80% (95% CI, 72.41-88.28%) in the detection of maternal cancer among pregnant women with MCA results. The sensitivity of the classifier was 93.33, 66.67, and 50%, while specificity was 66.67, 90, and 97.06%, and positive predictive value (PPV) was 60.87, 72.73, and 80% for the prediction of breast cancer, liver cancer, and lymphoma, respectively. This study presents a solution to identify maternal cancer patients from pregnant women with MCA results using NIPT, indicating it as a value-added application of NIPT in the detection of maternal malignancies in addition to screening for fetal aneuploidies with no extra cost.
母亲患恶性肿瘤可能导致无创产前检测（NIPT）出现假阳性结果或检测失败。尽管最近的研究表明，多种染色体非整倍体（MCA）与恶性肿瘤有关，但使用NIPT从出现MCA结果的孕妇中识别出患癌孕妇，仍然没有有效的解决办法。我们旨在开发一种新方法，利用NIPT有效地检测出出现MCA结果的孕妇中的母亲癌症，并使用随机森林分类器来识别常见母亲癌症类型的组织来源。为进行检测，2016年1月至2019年6月期间，在华大基因招募了496名通过NIPT获得MCA结果的参与者。通过临床随访确认癌症和非癌症参与者。利用2016年1月至2017年12月招募的42例母亲癌症病例和294例非癌症病例组成的队列，开发了一种名为前五位染色体z值均值（MTOP5Zscores）的方法。2018年1月至2019年6月招募的其余160名参与者用于验证MTOP5Zscores的性能。我们建立了一个随机森林模型，使用标准化皮尔逊相关系数（NPCC）值、22条染色体的z值和7种血浆肿瘤标志物（PTM）作为预测变量，对三种常见癌症类型进行分类。62例母亲癌症病例被确诊为乳腺癌、肝癌和淋巴瘤，这是最常见的癌症类型。MTOP5Zscores在检测出现MCA结果的孕妇中的母亲癌症时，灵敏度为85%（95%置信区间（CI），62.11 - 96.79%），特异性为80%（95%CI，72.41 - 88.28%）。该分类器对乳腺癌、肝癌和淋巴瘤预测时的灵敏度分别为93.33%、66.67%和50%，特异性分别为66.67%、90%和97.06%，阳性预测值（PPV）分别为60.87%、72.73%和80%。本研究提出了一种利用NIPT从出现MCA结果的孕妇中识别患癌孕妇的解决方案，表明这是NIPT在检测母亲恶性肿瘤方面的一项增值应用，除了筛查胎儿非整倍体之外无需额外费用。
