DNA Methylation Markers for Pan-Cancer Prediction by Deep Learning

Abstract

For cancer diagnosis, many DNA methylation markers have been identified. However, few studies have tried to identify DNA methylation markers to diagnose diverse cancer types simultaneously, i.e., pan-cancers. In this study, we tried to identify DNA methylation markers to differentiate cancer samples from the respective normal samples in pan-cancers. We collected whole genome methylation data of 27 cancer types containing 10,140 cancer samples and 3386 normal samples, and divided all samples into five data sets, including one training data set, one validation data set and three test data sets. We applied machine learning to identify DNA methylation markers, and specifically, we constructed diagnostic prediction models by deep learning. We identified two categories of markers: 12 CpG markers and 13 promoter markers. Three of 12 CpG markers and four of 13 promoter markers locate at cancer-related genes. With the CpG markers, our model achieved an average sensitivity and specificity on test data sets as 92.8% and 90.1%, respectively. For promoter markers, the average sensitivity and specificity on test data sets were 89.8% and 81.1%, respectively. Furthermore, in cell-free DNA methylation data of 163 prostate cancer samples, the CpG markers achieved the sensitivity as 100%, and the promoter markers achieved 92%. For both marker types, the specificity of normal whole blood was 100%. To conclude, we identified methylation markers to diagnose pan-cancers, which might be applied to liquid biopsy of cancers.

MethylNet: an automated and modular deep learning approach for DNA methylation analysis

“…Prior works that have explored pan-cancer prediction in the deep learning space have limited their analyses to a small set of CpGs that do not capture a holistic understanding of interaction and regulation in the cancer context [44]. Our results demonstrate that models with a larger number of CpGs are needed to accurately capture differences in tissue/cancer subtypes.…”

<p>MBD2 Correlates with a Poor Prognosis and Tumor Progression in Renal Cell Carcinoma</p>

“…Aberrant DNA methylation modifications are frequently detected in various tumors, and the main mechanisms for DNA methylation-involved tumorigenesis are that methylation levels of the promoter region of anti-oncogene are elevated, which promotes the key anti-oncogene silencing and drives tumorigenesis. 12,[14][15][16] MBD2 has been reported in multiple human malignancies, including gastric cancer, 17 breast cancer, 18,19 colorectal cancer, 20 glioblastoma, 21,22 hilar cholangiocarcinoma, 23 hepatocellular carcinoma, 24,25 chronic myeloid leukemia 26 and prostate cancer. 27 Previous studies confirmed that MBD2 mediates the transcriptional repression of tumor suppressor genes, such as hTERT, 28 GSTP1, 29 BAI1, 21 p14 ARF /p16 INK4a,30 and 14-3-3sigma, 27 which supports the pivotal role of MBD2 in abnormal epigenetic regulation of tumors.…”

Machine Classification of Methylomes in Cancer

“…Additionally, an accuracy of 0.961 is obtained by Tang et al, in multiclass classification of 14 tissue types [19]. Liu et al reaches a ROC AUC of 0.989 with multiclass classification of eight cancer types, where some samples are from TCGA [20]. These recent studies demonstrate that machine learning approaches can be used to classify cancer based on DNA methylation data, even when using a highly-filtered small probe list [21,20,19,18].…”

“…Liu et al reaches a ROC AUC of 0.989 with multiclass classification of eight cancer types, where some samples are from TCGA [20]. These recent studies demonstrate that machine learning approaches can be used to classify cancer based on DNA methylation data, even when using a highly-filtered small probe list [21,20,19,18]. Multiple probe selection statistical methods were commonly used, such as LASSO, the moderated t-statistic, Maximum-Relevance-Maximum-Distance, and PCA.…”