文献主要内容
以往基于液体活检早期检测进展期结直肠腺瘤(advCRA)或腺癌(CRC)的研究受到灵敏度低的限制。本文进行了一项前瞻性研究,利用cfDNA的片段组学谱(fragmentomic profles)建立一个整合模型,以准确和经济有效地检测早期CRC和advCRA。
文献思路
目前看到的几篇cfDNA的片段组学分析思路简直一模一样!我一度怀疑自己看了同一个团队的研究或者同一个研究,但是作者单位啥的又都不一样...
数据情况
- 训练集合:310个样本
- 测试集合:311个样本
进行全基因组测序,分析的片段组学特征有:Fragment Size Ratio (FSR)、Fragment Size Distribution (FSD)、6bp End Motif (EDM)、Breakpoint Motif (BPM)、Copy Number Variation (CNV)。
训练模型:ensembled stacked machine learning,包括5种算法:GLM、GBM、Random Forest、Deep learning、Xgboost
数据:此处分析也进行了降采样,WGS data were down sampled to 4X coverage
模型性能
- The ensembled stacked model:分类性能AUC值>0.988,而五个特征之一单一构建的模型分类性能AUC range 0.881–0.981。
- 检测advCRA/early-stage CRC 的sensitivity:The ensembled stacked model为97.4%, 95% CI 94.1–99.2%
- 6bp EDM比4bpEDM的好:the adapted 6-bp EDM feature showed higher AUC (0.981, 95% CI 0.969–0.993) than the original 4-bp end-motif feature
- FSR or FSD比original DELFI fragment pattern的好:models using FSR or FSD both had higher AUC (0.881, 95% CI 0.843–0.919; 0.892, 95% CI 0.855–0.930) than the original DELFI fragment pattern
每种指标单独建模性能结果:
不同测序深度下的模型结果:
数据分析方法
- cfDNA extraction:QIAamp Circulating Nucleic Acid Kit (Qiagen)
- cfDNA concentrations:Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific)
- WGS文库构建:the KAPA Hyper Prep Kit (KAPA Biosystems),5-10 ng of cfDNA;end-repairing, A-tailing, and ligation with adapters
- 文库定量:the KAPA SYBR FAST qPCR Master Mix (KAPA Biosystems)
- 测序:双端,Illumina NovaSeq平台
- 去接头:Trimmomatic
- 比对:GRCh37/UCSC hg19、bwa
- PCR去重:Picard toolkit (http://broadinstitute.github.io/picard/)
- 最终数据的测序深度: 4.7X to 24.04X (median coverage 9.75X)
- Fragment Size Ratio (FSR):与 DELFI定义类似,the short, intermediate and long fragments were defined as 65-150bp, 151-220bp and 221-400bp, according to the overall fragment lengths profile
- CNV:ichorCNA包
- ROC曲线:pROC package (v. 1.17.0.1)
- 敏感性、特异性等指标计算: epiR package (v 2.0.19)
- 年龄与性别的倾向得分匹配分析:MatchIt package (4 4.2.0)
文献信息
J Hematol Oncol. 2021 Oct 26;14(1):175. doi: 10.1186/s13045-021-01189-w.
Multi-dimensional fragmentomic assay for ultrasensitive early detection of colorectal advanced adenoma and adenocarcinoma