Hsu Yin-Chen, Huang Sin-Ming, Chang Li-Chun, Chen Yan-Ming, Chang Ya-Hsuan, Lin Jing-Wei, Lin Chien-Chia, Chen Ching-Wen, Chen Hsuan-Yu, Chiu Han-Mo, Yu Sung-Liang
Department of Clinical Laboratory Sciences and Medical Biotechnology, College of Medicine, National Taiwan University Taipei, Taiwan.
Department of Internal Medicine, National Taiwan University Hospital Taipei, Taiwan.
Am J Cancer Res. 2022 Mar 15;12(3):1088-1101. eCollection 2022.
Liquid biopsy test has a better uptake for colorectal cancer (CRC) screening. However, suboptimal detection of early-staged colorectal neoplasia (CRN) limits its application. Here, we established an early-staged CRN blood test using error-corrected sequencing by comparing clonal hematopoiesis (CH) of 63 CRN patients and that of 32 controls. We identified 1,446 variants and classified the uniqueness in CRN patients. There was no significance difference in the amount of variant between CRNs and controls, but the uniqueness of variants with defective DNA mismatch repair-related mutational signature was addressed from peripheral blood in early-staged CRN patients. By machine learning approach, the early-staged CRNs was discriminated from controls with an AUC of 0.959 and an accuracy of 0.937 (95% CI, 0.863 to 0.968). The CRN predictive model was further validated by additional 20 CRNs and 10 controls and showed the accuracy, sensitivity, specificity, positive prediction value (PPV) and negative prediction value (NPV) of 0.933 (95% CI: 0.779 to 0.992), 0.95, 0.90, 0.95 and 0.90, respectively. In summary, we develop a CH-based liquid biopsy test with machine learning approach, which not only increase screening uptake but also improve the detection rate of early-staged CRN.
液体活检检测在结直肠癌（CRC）筛查中的接受度更高。然而，早期结直肠肿瘤（CRN）的检测效果欠佳限制了其应用。在此，我们通过比较63例CRN患者和32例对照的克隆性造血（CH），利用纠错测序建立了一种早期CRN血液检测方法。我们鉴定出1446个变异，并对CRN患者中的变异独特性进行了分类。CRN患者和对照之间的变异数量没有显著差异，但早期CRN患者外周血中具有缺陷的DNA错配修复相关突变特征的变异独特性得到了体现。通过机器学习方法，早期CRN与对照得以区分，曲线下面积（AUC）为0.959，准确率为0.937（95%置信区间，0.863至0.968）。CRN预测模型通过另外20例CRN患者和10例对照进一步验证，其准确率、灵敏度、特异性、阳性预测值（PPV）和阴性预测值（NPV）分别为0.933（95%置信区间：0.779至0.992）、0.95、0.90、0.95和0.90。总之，我们利用机器学习方法开发了一种基于CH的液体活检检测方法，该方法不仅提高了筛查接受度，还提高了早期CRN的检测率。
