Additional file 1 of Integrated epigenetic biomarkers in circulating cell-free DNA as a robust classifier for pancreatic cancer

Additional file 1: Supplementary Figure 1. Percentage of reads mapped to the spike-in DNA. A. The 5mC spike-in DNA is specifically enriched in the 5mC libraries. B. The 5hmC spike-in DNA is specifically enriched in the 5hmC libraries. Error bars indicate Standard Deviation (SD). Supplementary Figure 2. Global change in 5mC and 5hmC level in PDAC. A. Boxplot of 5mC peak numbers from healthy controls and PDAC samples shows no significant difference. B. Boxplot of 5hmC peak numbers from healthy controls and PDAC samples shows significant larger numbers of peaks in PDAC samples. C. Genome browser view of the cell-free 5mC distribution in a 5 mb region in chromosome 8. D. Genome browser view of the cell-free 5hmC distribution in a 3 mb region in chromosome 7. The overlapping tracks of healthy and PDAC are shown in line plot. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 1e−5, Wilcoxon test. PDAC, pancreatic ductal adenocarcinoma. Supplementary Figure 3. Genomic distribution of 5mC and 5hmC peaks. A. 5mC distribution in genomic features. B. Enrichment of 5mC peaks overlapping with distinct genomic elements. C. 5hmC peak distribution in genomic features. D. Enrichment of 5hmC peaks overlapping with distinct genomic elements. PDAC, pancreatic ductal adenocarcinoma; CDS, Coding DNA Sequence; 3′UTR, 3′untranslated region; 5′UTR, 5′untranslated region. Supplementary Figure 4. Comparison of the 5mC and 5hmC peaks. A. Venn diagram of overlap between 5mC and 5hmC peaks. B. Venn diagram of overlap between genes with 5mC modifications and genes with 5hmC modifications. Supplementary Figure 5. GO term enrichment analysis of specifically modified genes. A. 5mC-specific genes. B. 5hmC-specific genes. PDAC, pancreatic ductal adenocarcinoma. Supplementary Figure 6. Genome browser views of examples of specifically modified genes. A. ME1 gene in chromosome 6: 84,095–84,140 kb. B. PACRG gene in chromosome 6: 163,716–163,734 kb. C. FYN gene in chromosome 6: 112,132–112,148 kb. D. RALB gene in chromosome 2: 121,000–121,030 kb. ME1, malic enzyme 1; PACRG, parkin coregulated; FYN, FYN proto-oncogene; RALB, RAS like proto-oncogene B. Supplementary Figure 7. T-SNE analysis of 5mC FPKM. A. T-SNE plot of 5mC FPKM from PDAC and healthy samples in distinct batches after removing batch effect. B. T-SNE plot of 5mC FPKM from PDAC samples and healthy samples. PDAC, pancreatic ductal adenocarcinoma. Supplementary Figure 8. Flow chart of 5mC model construction. Supplementary Figure 9. T-SNE analysis of 5hmC FPKM. A. T-SNE plot of 5hmC FPKM from PDAC and healthy samples in distinct batches. B. T-SNE plot of 5hmC FPKM from PDAC and healthy samples. PDAC, pancreatic ductal adenocarcinoma. Supplementary Figure 10. Flow chart of 5hmC model construction. Supplementary Figure 11. Performance of the 5hmC model in distinguishing the subgroups of PDAC patients. A. Boxplot of the wd-scores in the resectable PDAC patients and unresectable PDAC patients. B. Boxplot of the wd-scores in PDAC patients with jaundice and those without jaundice. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 1e−5, Wilcoxon test.

附加文件1：补充图1。比对至外参DNA（spike-in DNA）的reads占比。A. 5-甲基胞嘧啶DNA（5mC spike-in DNA）在5mC文库中特异性富集。B. 5-羟甲基胞嘧啶DNA（5hmC spike-in DNA）在5hmC文库中特异性富集。误差棒表示标准差（Standard Deviation, SD）。
补充图2。胰腺导管腺癌（pancreatic ductal adenocarcinoma, PDAC）中5mC与5hmC水平的全局变化。A. 健康对照与PDAC样本的5mC峰数箱线图显示无显著差异。B. 健康对照与PDAC样本的5hmC峰数箱线图显示PDAC样本的峰数显著更多。C. 8号染色体5 Mb区域内无细胞5mC分布的基因组浏览器视图。D. 7号染色体3 Mb区域内无细胞5hmC分布的基因组浏览器视图。健康对照与PDAC的叠加轨迹以折线图展示。*P < 0.05，**P < 0.01，***P < 0.001，****P < 1e−5，威尔科克森检验（Wilcoxon test）。PDAC，胰腺导管腺癌（pancreatic ductal adenocarcinoma）。
补充图3。5mC与5hmC峰的基因组分布。A. 5mC在基因组特征中的分布。B. 5mC峰与不同基因组元件重叠的富集情况。C. 5hmC峰在基因组特征中的分布。D. 5hmC峰与不同基因组元件重叠的富集情况。PDAC，胰腺导管腺癌（pancreatic ductal adenocarcinoma）；CDS，编码序列（Coding DNA Sequence, CDS）；3′UTR，3'非翻译区（3′untranslated region, 3′UTR）；5′UTR，5'非翻译区（5′untranslated region, 5′UTR）。
补充图4。5mC与5hmC峰的比较。A. 5mC与5hmC峰重叠的维恩图。B. 携带5mC修饰的基因与携带5hmC修饰的基因重叠的维恩图。
补充图5。特异性修饰基因的GO术语富集分析。A. 5mC特异性修饰基因。B. 5hmC特异性修饰基因。PDAC，胰腺导管腺癌（pancreatic ductal adenocarcinoma）。
补充图6。特异性修饰基因示例的基因组浏览器视图。A. 6号染色体上的ME1基因：84,095–84,140 kb。B. 6号染色体上的PACRG基因：163,716–163,734 kb。C. 6号染色体上的FYN基因：112,132–112,148 kb。D. 2号染色体上的RALB基因：121,000–121,030 kb。ME1，苹果酸酶1（malic enzyme 1, ME1）；PACRG，帕金共调节因子（parkin coregulated, PACRG）；FYN，FYN原癌基因（FYN proto-oncogene, FYN）；RALB，RAS样原癌基因B（RAS like proto-oncogene B, RALB）。
补充图7。5mC FPKM的t分布邻域嵌入（T-SNE）分析。A. 去除批次效应后，不同批次PDAC与健康样本的5mC FPKM的T-SNE图。B. PDAC与健康样本的5mC FPKM的T-SNE图。PDAC，胰腺导管腺癌（pancreatic ductal adenocarcinoma）。
补充图8。5mC模型构建流程图。
补充图9。5hmC FPKM的T-SNE分析。A. 不同批次PDAC与健康样本的5hmC FPKM的T-SNE图。B. PDAC与健康样本的5hmC FPKM的T-SNE图。PDAC，胰腺导管腺癌（pancreatic ductal adenocarcinoma）。
补充图10。5hmC模型构建流程图。
补充图11。5hmC模型区分PDAC患者亚组的性能。A. 可切除PDAC患者与不可切除PDAC患者的wd-scores箱线图。B. 伴黄疸与不伴黄疸PDAC患者的wd-scores箱线图。*P < 0.05，**P < 0.01，***P < 0.001，****P < 1e−5，威尔科克森检验（Wilcoxon test）。